Solar sound and light device group cooperative control optimization method and system based on internet of things

Abstract

This invention provides a collaborative control optimization method and system for a group of solar-powered acoustic and optical devices based on the Internet of Things (IoT), relating to the field of device control technology. The method includes analyzing the correlation of device data sequence features using a Long Short-Term Memory (LSTM) network, calculating device trust values ​​based on temporal changes in device feature identifiers, dividing the device group into a core node layer and a collaborative node layer, establishing a verification sequence using a Bloom filter, and forming a cross-domain resource allocation scheme by combining the LSTM network. This invention improves the collaborative control efficiency of a group of solar-powered acoustic and optical devices and enhances system security and adaptability.

Description

Technical Field

[0001] This invention relates to equipment control technology, and more particularly to a method and system for collaborative control optimization of a group of solar-powered acoustic and optical equipment based on the Internet of Things. Background Technology

[0002] With the rapid development of IoT technology and the widespread application of renewable energy, solar-powered audio-visual equipment has become an important component in smart cities, intelligent transportation, and other scenarios. Solar-powered audio-visual equipment clusters typically consist of multiple distributed nodes. These devices are interconnected through IoT technology to jointly perform specific monitoring, warning, or information dissemination functions. Due to their clean, environmentally friendly, and self-sufficient energy supply characteristics, solar-powered audio-visual equipment has been widely used in scenarios such as road traffic, border security, and disaster early warning. Currently, the control and management of solar-powered audio-visual equipment clusters mainly rely on centralized control systems, which use preset control strategies to uniformly schedule and manage the equipment.

[0003] Existing technologies lack in-depth mining and utilization of historical equipment operation data, making it impossible to effectively analyze the temporal characteristics and correlations of equipment operation status. This results in control strategies lacking pertinence and foresight, and failing to perform precise control and scheduling optimization based on the equipment's historical performance.

[0004] Traditional control methods typically employ a uniform control strategy to manage all devices, without considering the differences and hierarchical nature between devices. They fail to establish a hierarchical control mechanism based on device characteristics and trust levels, making it difficult to achieve differentiated collaboration and optimized resource allocation within device groups.

[0005] Existing collaborative systems for solar-powered acoustic and optical equipment groups lack secure and reliable mechanisms for cross-domain data interaction and resource allocation. The security and integrity of data transmission are difficult to guarantee. Furthermore, there is a lack of intelligent allocation methods based on equipment characteristics and time-series correlations, which limits the collaborative efficiency and adaptability of equipment groups in complex environments. Summary of the Invention

[0006] This invention provides an IoT-based method for the collaborative control and optimization of a group of solar-powered acoustic and optical devices, which can solve the problems in the prior art.

[0007] A first aspect of the present invention provides a collaborative control optimization method for a group of solar-powered acoustic and optical devices based on the Internet of Things, comprising:

[0008] The positional source data of the solar-powered acoustic and optical equipment group is mapped into multi-dimensional temporal features to construct an equipment data sequence; the long short-term memory network is used to analyze the feature correlation in the equipment data sequence, quantify the historical operating characteristics of the equipment, and mark the equipment feature identifiers;

[0009] The device trust value is calculated based on the temporal change trend of the device feature identifier, and the solar-powered acoustic and optical device group is divided into a core node layer and a cooperative node layer according to the distribution characteristics of the device trust value.

[0010] Based on the Long Short-Term Memory network, the control command sequence of the core node layer is decoupled, and a mapping relationship between the device feature identifier and the control command is established to generate an initial control scheme. According to the Bloom filter of the cooperative node layer, a verification sequence of the initial control scheme is established, and the device trust value is updated by matching the device feature identifier to form the final control scheme.

[0011] Based on the control sequence formed by the final control scheme, the intelligent control status of the solar acoustic and optical equipment group is analyzed in real time to construct cross-domain collaborative data. According to the Bloom filter, a hash mapping is established on the cross-domain collaborative data to generate security data. The long short-term memory network is used to analyze the temporal characteristics of the security data and establish an association model with the device feature identifier to form a cross-domain resource allocation scheme. The cross-domain resource allocation scheme is converted into device execution instructions through control sequence mapping.

[0012] The positional source data of the solar-powered acoustic and optical equipment group is mapped into multi-dimensional temporal features to construct an equipment data sequence. A long short-term memory network is used to analyze the feature correlation in the equipment data sequence, quantify the historical operating characteristics of the equipment, and label the equipment features, including:

[0013] The positional source data of the solar acousto-optic device group is constructed into a time-series data sequence, a quantum well-quantum dot composite structure is constructed, the carrier transport path is controlled based on the quantum confinement effect of the quantum well-quantum dot composite structure, the time-series data sequence is mapped into quantum states, and the energy level distribution of the quantum well-quantum dot composite structure is adjusted through quantum state engineering to form a quantum feature sequence.

[0014] A quantum transmission channel is constructed using the quantum tunneling effect. The quantum transmission channel includes a quantum tunneling barrier array. The carrier transport is controlled by the Coulomb blocking effect of the quantum tunneling barrier array. A device data sequence is generated based on the coupling effect of the quantum tunneling effect and the quantum confinement effect.

[0015] A long short-term memory network is established to parse the device data sequence. The output features of the long short-term memory network are correlated and mapped with the coherence characteristics of the quantum state. Based on the coherence function of the quantum state, the historical operating characteristics of the device are quantized to generate device feature identifiers.

[0016] A quantum transmission channel is constructed using the quantum tunneling effect. The quantum transmission channel includes a quantum tunneling barrier array. Carrier transport is modulated through the Coulomb blockade effect of the quantum tunneling barrier array. A device data sequence is generated based on the coupling effect of the quantum tunneling effect and the quantum confinement effect, including:

[0017] A periodically arranged quantum tunneling barrier array is constructed, and a gradient micro / nano structure is used to establish a carrier wave function in the barrier region of the quantum tunneling barrier array. The exponential wave vector parameter of the carrier wave function is modulated based on the gradient micro / nano structure.

[0018] A Coulomb blocking effect control mechanism is constructed under the action of the current carrier wavefunction. The transmission coefficient is generated by enhancing the coupling effect of the left and right potential barriers using surface plasmon resonance. The tunneling current corresponding to the current carrier wavefunction is controlled based on the transmission coefficient.

[0019] The Coulomb blocking effect control mechanism and the quantum confinement effect are coupled into a system. The coupled state transmission and uncoupled state transmission of the coupled system are realized by external field control, generating a coupling enhancement factor. The tunneling current is modulated according to the coupling enhancement factor to generate a device data sequence.

[0020] Based on the Long Short-Term Memory network, the control command sequence of the core node layer is decoupled, and a mapping relationship between the device feature identifier and the control command is established to generate an initial control scheme, including:

[0021] A long short-term memory network is trained to establish a temporal correlation model of device features. Based on the temporal correlation model, the control command sequence of the core node layer is decoupled. A gradient feedback adjustment mechanism is used to construct a mapping relationship between the device feature identifier and the control command. The gradient feedback adjustment mechanism adjusts the mapping weights in real time. An initial control scheme is generated based on the iterative optimization results of the mapping weights.

[0022] The initial control scheme is established based on the Bloom filter of the cooperative node layer, and the device trust value is updated by matching the device feature identifier to form the final control scheme, which includes:

[0023] The initial control scheme verification sequence is constructed based on the Bloom filter of the collaborative node layer. The initial control scheme is decomposed into multiple control scheme components. The control scheme components are mapped using a family of hash functions. The mapping results are input into the neural symbolic reasoning system. Based on the neural symbolic reasoning system and the expert knowledge base, an optimized verification sequence is generated.

[0024] A causal inference-driven feature matching network is constructed based on the optimized verification sequence. Causal correlation analysis is performed on the feature components and the optimized verification sequence based on the causal inference-driven feature matching network. The device trust value is calculated using a multi-agent collaborative evaluation mechanism.

[0025] The device trust value is input into the transfer learning model, which generates an optimized control strategy based on the basic control strategy and new scenario features. The optimized control strategy is then mapped to a trust level according to the device trust value to generate the final control scheme.

[0026] The control scheme components are mapped using a family of hash functions, and the mapping results are input into a neural symbolic reasoning system. Based on the neural symbolic reasoning system and an expert knowledge base, an optimized verification sequence is generated, including:

[0027] The control scheme is decomposed into multiple control components. The multiple control components are mapped using a family of hash functions to obtain initial mapping features. A component feature matrix is ​​constructed based on the initial mapping features. The degree of correlation between components is determined by similarity calculation. The component feature matrix is ​​adjusted according to the degree of correlation to generate enhanced mapping features.

[0028] The enhanced mapping features are input into the neural symbolic reasoning system. The neural symbolic reasoning system extracts information from the enhanced mapping features to construct a feature priority sequence. The enhanced mapping features are reconstructed based on the feature priority sequence to obtain reconstructed features. The temporal variation pattern of the reconstructed features is analyzed to generate temporal correlation features.

[0029] A knowledge fusion module is constructed, and a mapping relationship between the temporal correlation features and the expert knowledge base is established based on the knowledge fusion module. A feature expression space is constructed according to the mapping relationship, and the temporal correlation features are transformed into standard feature representations in the feature expression space. An optimized verification sequence is generated by adjusting the standard feature representations based on a balance factor.

[0030] Based on the control sequence formed by the final control scheme, the intelligent control status of the solar-powered acoustic and optical equipment group is analyzed in real time to construct cross-domain collaborative data; security data is generated by establishing a hash mapping on the cross-domain collaborative data according to the Bloom filter, including:

[0031] The control sequence is chaotically mapped to obtain the device state iteration value. A chaotic mapping model is constructed based on the device state iteration value. The dynamic change law of the device state parameters is analyzed based on the chaotic mapping model. A parameter mapping sequence is generated based on the dynamic change law.

[0032] The state change trend of the device group is analyzed according to the parameter mapping sequence. The state change trend reflects the collaborative evolution process of the device group under the action of the parameter mapping sequence. The operational stability of the device group is judged based on the collaborative evolution process. The control strategy is dynamically adjusted according to the operational stability using the chaotic mapping model.

[0033] The control strategy is constructed using the chaotic mapping model, a hash verification space is established in the hash verification space, and the security verification information of the control strategy is transmitted along the verification path to generate security data.

[0034] A second aspect of the present invention provides an Internet of Things-based collaborative control and optimization system for a group of solar-powered acoustic and optical devices, comprising:

[0035] The first unit is used to map the positional source data of the solar-powered acoustic and optical equipment group into multi-dimensional temporal features and construct the equipment data sequence; it uses a long short-term memory network to analyze the feature correlation in the equipment data sequence, quantifies the historical operating features of the equipment, and marks the equipment feature identifiers.

[0036] The second unit is used to calculate the device trust value based on the temporal change trend of the device feature identifier, and to divide the solar-powered acoustic and optical device group into a core node layer and a cooperative node layer according to the distribution characteristics of the device trust value.

[0037] The third unit is used to decouple the control command sequence of the core node layer based on the long short-term memory network, establish the mapping relationship between the device feature identifier and the control command, and generate an initial control scheme; establish a verification sequence of the initial control scheme according to the Bloom filter of the cooperative node layer, match the device feature identifier to update the device trust value, and form a final control scheme.

[0038] The fourth unit is used to analyze the intelligent control status of the solar-powered acoustic and optical equipment group in real time based on the control sequence formed by the final control scheme, and construct cross-domain collaborative data; to establish a hash mapping of the cross-domain collaborative data according to the Bloom filter to generate security data, to use the Long Short-Term Memory Network to analyze the temporal characteristics of the security data and establish an association model with the device feature identifier, to form a cross-domain resource allocation scheme, and the cross-domain resource allocation scheme is converted into device execution instructions through control sequence mapping.

[0039] A third aspect of the present invention provides an electronic device, comprising:

[0040] processor;

[0041] Memory used to store processor-executable instructions;

[0042] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0043] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0044] The beneficial effects of this application are as follows:

[0045] By mapping the positional source data of a solar-powered acoustic-optical device group into multi-dimensional temporal features and using a long short-term memory network to analyze the feature correlation in the device data sequence, the accurate quantification and labeling of the historical operating features of the devices were achieved, improving the accuracy and timeliness of device feature identification and laying a data foundation for subsequent collaborative control.

[0046] By using device trust value calculation and node hierarchy division mechanism, a hierarchical control architecture of core node layer and cooperative node layer was established, realizing precise decoupling and mapping of control commands. At the same time, Bloom filter was used to establish verification sequence for control scheme optimization, effectively improving the collaborative control efficiency and reliability of solar acoustic and optical equipment group.

[0047] By constructing cross-domain collaborative data and using Bloom filters to establish hash mappings to generate secure data, and combining long short-term memory networks to analyze timing characteristics, secure and efficient communication and resource optimization among device groups are achieved, significantly enhancing the system's security, stability, and adaptability. This solves the problems of low collaborative efficiency and insufficient security of traditional control methods in complex environments. Attached Figure Description

[0048] Figure 1 This is a flowchart illustrating the collaborative control optimization method for a group of solar-powered acoustic and optical devices based on the Internet of Things, according to an embodiment of the present invention.

[0049] Figure 2 This is a flowchart illustrating the construction process of a quantum transmission channel according to an embodiment of the present invention.

[0050] Figure 3 The flowchart for generating optimized verification sequence logic for the neural symbolic reasoning system in this embodiment of the invention is shown below. Detailed Implementation

[0051] 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, and 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.

[0052] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0053] Figure 1This is a flowchart illustrating the collaborative control optimization method for a group of solar-powered acoustic and optical devices based on the Internet of Things, as described in an embodiment of the present invention. Figure 1 As shown, the method includes:

[0054] The positional source data of the solar-powered acoustic and optical equipment group is mapped into multi-dimensional temporal features to construct an equipment data sequence; the long short-term memory network is used to analyze the feature correlation in the equipment data sequence, quantify the historical operating characteristics of the equipment, and mark the equipment feature identifiers;

[0055] The device trust value is calculated based on the temporal change trend of the device feature identifier, and the solar-powered acoustic and optical device group is divided into a core node layer and a cooperative node layer according to the distribution characteristics of the device trust value.

[0056] Based on the Long Short-Term Memory network, the control command sequence of the core node layer is decoupled, and a mapping relationship between the device feature identifier and the control command is established to generate an initial control scheme. According to the Bloom filter of the cooperative node layer, a verification sequence of the initial control scheme is established, and the device trust value is updated by matching the device feature identifier to form the final control scheme.

[0057] Based on the control sequence formed by the final control scheme, the intelligent control status of the solar acoustic and optical equipment group is analyzed in real time to construct cross-domain collaborative data. According to the Bloom filter, a hash mapping is established on the cross-domain collaborative data to generate security data. The long short-term memory network is used to analyze the temporal characteristics of the security data and establish an association model with the device feature identifier to form a cross-domain resource allocation scheme. The cross-domain resource allocation scheme is converted into device execution instructions through control sequence mapping.

[0058] In one optional implementation, the positional source data of the solar-powered acoustic and optical device group is mapped into multi-dimensional temporal features to construct a device data sequence; a long short-term memory network is used to analyze the feature correlation in the device data sequence, quantify the historical operating characteristics of the devices, and label the device feature identifiers, including:

[0059] The positional source data of the solar acousto-optic device group is constructed into a time-series data sequence, a quantum well-quantum dot composite structure is constructed, the carrier transport path is controlled based on the quantum confinement effect of the quantum well-quantum dot composite structure, the time-series data sequence is mapped into quantum states, and the energy level distribution of the quantum well-quantum dot composite structure is adjusted through quantum state engineering to form a quantum feature sequence.

[0060] A quantum transmission channel is constructed using the quantum tunneling effect. The quantum transmission channel includes a quantum tunneling barrier array. The carrier transport is controlled by the Coulomb blocking effect of the quantum tunneling barrier array. A device data sequence is generated based on the coupling effect of the quantum tunneling effect and the quantum confinement effect.

[0061] A long short-term memory network is established to parse the device data sequence. The output features of the long short-term memory network are correlated and mapped with the coherence characteristics of the quantum state. Based on the coherence function of the quantum state, the historical operating characteristics of the device are quantized to generate device feature identifiers.

[0062] The acquisition of positional source data for the solar-powered acoustic-optical device cluster was accomplished through a distributed sensor network. The acquisition system consisted of 48 nodes, each equipped with a power sensor, angle sensor, illumination sensor, and acoustic sensor, with a sampling frequency set at 100 Hz. The positional source data included device power data (range 0-500 watts, accuracy 0.1 watts), orientation angle data (range 0-360 degrees, accuracy 0.5 degrees), illumination intensity data (range 0-2000 lumens, accuracy 1 lumen), and acoustic output data (range 30-100 dB, accuracy 0.2 dB). The data acquisition time window was set to 24 hours, with each device generating 345,600 sampling points, resulting in a total of 16,588,800 raw data points for the entire cluster. The raw data underwent preprocessing via an edge computing unit, including outlier filtering (values ​​outside the normal range were marked as invalid), missing value imputation (using linear interpolation), and data standardization (mapping to the range of -1 to 1). The processed positional source data is organized into a time-series data sequence, with each device forming an 8-dimensional time-series vector (power, power change rate, angle, angle change rate, illumination intensity, illumination change rate, acoustic output, acoustic output change rate), and the sequence length is 86,400 points (downsampled at 10 Hz).

[0063] The quantum well-quantum dot composite structure was constructed using molecular beam epitaxy (MBE). A 10 nm Al0.3Ga0.7As barrier layer, a 12 nm GaAs quantum well layer, a 2 nm Al0.3Ga0.7As spacer layer, and a 4 nm In0.25Ga0.75As quantum dot layer (dot density 8 × 102) were sequentially grown on a GaAs substrate. 10 centimeter -2The structure consists of an average diameter of 18 nm and a 10 nm Al0.3Ga0.7As capping layer. Growth was controlled at 580°C at a growth rate of 0.8 nm / min. The quantum well layer provides a two-dimensional confinement space with a level spacing of 45 mEV; the quantum dot layer provides a zero-dimensional confinement space with a ground state to first excited state level spacing of 78 mEV. Carrier transport path modulation is achieved through an external electric field. 30 nm Ti / Au electrodes are deposited at the top and bottom of the structure as the top and back gates, respectively. The top gate voltage ranges from -1.0 to +1.0 V, and the back gate voltage ranges from -5.0 to +5.0 V. When the top gate voltage is +0.5 volts and the back gate voltage is +2.0 volts, the carrier wavefunctions in the quantum well and the quantum dot overlap to the maximum, forming an optimal transmission channel, and the measured tunneling current is 125 nanoamps; when the top gate voltage is -0.5 volts and the back gate voltage is -2.0 volts, the wavefunction overlap is minimal, and the tunneling current drops to 8 nanoamps.

[0064] The temporal data sequence vector state mapping adopts quantum encoding technology. Each 8-dimensional temporal vector is mapped to the quantum state of a 3-qubit system. The mapping relationship is as follows: the power parameter is mapped to the θ angle (0-π) of the first qubit, the angle parameter is mapped to the φ angle (0-2π) of the first qubit, the light intensity is mapped to the θ angle of the second qubit, the acoustic output is mapped to the φ angle of the second qubit, and the four rate of change parameters are mapped to the θ angle and φ angle of the third qubit and the two global phases, respectively. In the actual mapping, the parameters of device D1 at t=1000 seconds are as follows: power 380 watts (normalized value 0.52) mapped to θ1=1.63 radians, orientation angle 175 degrees (normalized value -0.03) mapped to φ1=3.08 radians, illuminance 650 lumens (normalized value 0.3) mapped to θ2=1.25 radians, acoustic output 65 dB (normalized value 0.2) mapped to φ2=3.77 radians, and the four rate-of-change parameters (0.05, -0.02, 0.03, 0.01) are mapped to θ3=1.58 radians, φ3=3.02 radians, and two global phases 0.03π and 0.01π, respectively. The quantum state obtained after mapping is verified by quantum state tomography with a fidelity of 97.5%.

[0065] Quantum state engineering manipulation employs microwave pulse technology, integrating a microwave resonant cavity on the surface of a quantum well-quantum dot composite structure. The resonant frequency is 8.5 GHz, and the quality factor Q = 12000. The microwave pulse sequence includes π pulses (20 nanoseconds pulse width, 15 milliwatts power) and π / 2 pulses (10 nanoseconds pulse width, 7.5 milliwatts power). Precise manipulation of the quantum state is achieved by modulating the phase, amplitude, and timing of the microwave pulses. The microwave pulse sequence is generated by an arbitrary waveform generator and transmitted to the resonant cavity via a coaxial cable. Energy level distribution manipulation is achieved by applying a perpendicular electric field with a strength range of 0-50 kV / cm and a step size of 1 kV / cm. At an electric field strength of 35 kV / cm, the energy levels of the quantum well and quantum dot achieve optimal matching, with the cross-coupling strength between energy levels reaching a maximum of 6.8 millielectronvolts. The formation of the quantum characteristic sequence is achieved through continuous measurement of the quantum state at 100-millisecond intervals, with 864,000 consecutive measurements to obtain the complete quantum characteristic sequence. The coherence time of the sequence is 1.5 microseconds, which is much longer than the single measurement time (10 nanoseconds), ensuring the validity of the measurement.

[0066] The quantum tunneling barrier array was constructed using selective epitaxy and precise etching techniques to grow a 10-layer periodic structure on a GaAs substrate. Each layer contains a 3.5 nm Al0.4Ga0.6As barrier layer and a 5.0 nm GaAs well layer. The barrier layer has a band gap of 0.32 eV, and the well layer forms a quantum confinement structure with a confinement energy of 78 mEV. Source and drain electrodes were fabricated at the top and bottom of the array, respectively, using Ti / Au materials (10 nm and 40 nm thick, respectively). Metal was deposited using electron beam evaporation at a vacuum level of 5 × 102 -7 The deposition rate was 0.5 Å / s. Gates were fabricated on the array sides using Cr / Au materials (5 nm and 30 nm thick, respectively). Barrier transmission characteristics were modulated by precisely controlling the gate voltage. The Coulomb blocking effect was achieved by introducing 20 nm diameter quantum dots in the fifth potential well. These quantum dots, made of InGaAs, were grown via molecular beam epitaxy at 520°C for 60 seconds. The quantum dot charging energy was 18 mEV, significantly higher than the thermal energy (approximately 25 mEV at room temperature), ensuring the effectiveness of the Coulomb blocking effect.

[0067] Carrier transport control is achieved by adjusting the gate voltage, which is set from -2.0 to +2.0 volts in 0.1-volt increments. At a gate voltage of -1.5 volts, the Coulomb blockade effect is strongest, and the tunneling current drops to a minimum of 5 nanoamps. At a gate voltage of +1.5 volts, the Coulomb blockade effect is suppressed, and the tunneling current reaches a maximum of 350 nanoamps. Periodic variations in the gate voltage generate a pulsed current sequence with a pulse width of 500 nanoseconds, a pulse interval of 1 microsecond, and a pulse height of 340 nanoamps. The coupling between the quantum tunneling effect and the quantum confinement effect is achieved by adjusting the relative positions of the quantum well-quantum dot composite structure and the quantum tunneling barrier array. The distance between the two structures is 8 nanometers, forming a coupled system through Coulomb interactions. The coupling strength is determined by measuring the enhancement factor of the tunneling current; the maximum enhancement factor is 5.8, corresponding to an increase in the tunneling current from 60 nanoamps to 348 nanoamps. The device data sequence is obtained by inputting a quantum feature sequence into the coupling system and measuring the output tunneling current sequence. The tunneling current sequence is then converted from analog to digital (12-bit resolution, 50 MHz sampling rate) to form the device data sequence, which has a length of 4,320,000 points.

[0068] The Long Short-Term Memory (LSTM) network was constructed using a deep learning framework. The network structure includes an input layer (256 neurons), three LSTM hidden layers (128, 64, and 32 neurons respectively), and an output layer (16 neurons). The input data consisted of device data sequences with a length of 1000 time steps, each containing a 256-dimensional feature vector. Network training used a batch size of 64, a learning rate of 0.001, and 200 training epochs. The training dataset contained 30 days of historical data, totaling 27,000 samples; the validation set contained 5 days of data, totaling 4,500 samples; and the test set contained 3 days of data, totaling 2,700 samples. Early stopping was used during training to avoid overfitting; training was stopped if the validation loss showed no improvement for 10 consecutive epochs. The final model achieved an accuracy of 94.8%, a recall of 92.6%, and an F1 score of 93.7% on the test set.

[0069] The correlation mapping between LSTM output features and quantum state coherence properties employs an attention mechanism. A mapping relationship is established between the 32-dimensional output features of the last hidden layer of the LSTM and the elements of the quantum state density matrix. A self-attention layer is introduced during the mapping process, containing four attention heads, each with an 8-dimensional dimension. The attention weights are determined through training, reflecting the correlation between different features and quantum state elements. Taking data from device D1 as an example, the attention mechanism identifies a high correlation between power fluctuation features and off-diagonal elements of the quantum state (correlation coefficient 0.87), and a high correlation between angular stability features and diagonal elements of the quantum state (correlation coefficient 0.92). Through this mapping relationship, a bidirectional conversion between classical data features and quantum state properties is achieved, with a conversion accuracy of 91.5%.

[0070] The quantum state coherence function is constructed based on density matrix representation. Each device's quantum state is represented by an 8×8 density matrix. The diagonal elements of the matrix reflect the operating state of each functional module within the device, while the off-diagonal elements reflect the cooperative relationships between modules. The coherence function is calculated using the purity and von Neumann entropy of the density matrix. The purity ranges from 0.125 to 1, and the entropy ranges from 0 to 3. Normal operating conditions correspond to a purity higher than 0.8 and an entropy lower than 0.7; abnormal conditions correspond to a purity lower than 0.6 and an entropy higher than 1.5. Taking device D5 as an example, its average purity calculated from 30 days of operation data is 0.87, and the average entropy is 0.52, indicating good operating conditions. In contrast, device D12 has an average purity of 0.58 and an average entropy of 1.74, indicating a malfunction.

[0071] The historical operational characteristics of the equipment were quantified using a sliding window technique, with a window size of 24 hours (86,400 data points) and a sliding step size of 1 hour (3,600 data points). Within each window, four statistical features of the quantum state coherence function were calculated: mean, variance, peak value, and trend. A total of 720 window feature vectors were generated over a 30-day operating period. These feature vectors underwent principal component analysis (PCA) dimensionality reduction, retaining the first eight principal components, explaining 94.3% of the variance. The dimensionality-reduced feature vectors were then used to classify the equipment status into three categories: "normal," "sub-healthy," and "faulty." The classification accuracy reached 96.7%, with a false positive rate of less than 2.5% and a false negative rate of less than 1.8%.

[0072] The device feature identifier generation employs hash encoding technology. The quantified historical operational feature vector of the device is input into the SHA-256 hash algorithm to generate a 256-bit hash value as the basis for the device feature identifier. This hash value is then encoded into a 64-character string, such as "7A3B5C2D9E8F4G6H". The first 16 characters of the feature identifier represent basic device information (model, location, etc.), the middle 32 characters represent historical operational features, and the last 16 characters represent current status information. The feature identifier is updated every 24 hours, retaining historical versions during updates to form a feature identifier chain. By comparing feature identifiers at adjacent time points, the degree of change in device status can be calculated; a change greater than 30% triggers an alarm mechanism. Practical application shows that this feature identifier technology can predict device failures 72 hours in advance with an accuracy rate of 89.5%, effectively reducing unexpected downtime and maintenance costs.

[0073] In one optional implementation, a quantum transmission channel is constructed using the quantum tunneling effect. The quantum transmission channel includes a quantum tunneling barrier array. Carrier transport is modulated through the Coulomb blocking effect of the quantum tunneling barrier array. A device data sequence is generated based on the coupling effect between the quantum tunneling effect and the quantum confinement effect, including:

[0074] A periodically arranged quantum tunneling barrier array is constructed, and a gradient micro / nano structure is used to establish a carrier wave function in the barrier region of the quantum tunneling barrier array. The exponential wave vector parameter of the carrier wave function is modulated based on the gradient micro / nano structure.

[0075] A Coulomb blocking effect control mechanism is constructed under the action of the current carrier wavefunction. The transmission coefficient is generated by enhancing the coupling effect of the left and right potential barriers using surface plasmon resonance. The tunneling current corresponding to the current carrier wavefunction is controlled based on the transmission coefficient.

[0076] The Coulomb blocking effect control mechanism and the quantum confinement effect are coupled into a system. The coupled state transmission and uncoupled state transmission of the coupled system are realized by external field control, generating a coupling enhancement factor. The tunneling current is modulated according to the coupling enhancement factor to generate a device data sequence.

[0077] like Figure 2 As shown, the method includes:

[0078] Prepare a silicon (100) crystal orientation substrate, and ultrasonically clean it sequentially with acetone, isopropanol, and deionized water for 10 minutes each. After drying with nitrogen, anneal it in a 580°C annealing furnace for 30 minutes to remove the surface oxide layer. Use a molecular beam epitaxy apparatus in an ultra-high vacuum environment (1×10⁻⁶). -10 First, a 2 nm gallium arsenide (GaAs) buffer layer was grown, followed by alternating growth of an aluminum GaAs barrier layer and a GaAs well layer. The growth parameters for the aluminum GaAs barrier layer were: temperature 580°C, V / III ratio 20:1, growth rate 1.0 nm / min, and growth time 210 s to form a 3.5 nm thickness. The growth parameters for the GaAs well layer were: temperature 600°C, V / III ratio 15:1, growth rate 1.2 nm / min, and growth time 250 s to form a 5.0 nm thickness. The growth of the barrier and well layers was repeated 10 times to form a periodically arranged quantum tunneling barrier array. After growth, the array was annealed at 185°C for 15 minutes to smooth the interface. X-ray diffraction was used to measure the periodicity, confirming that the periodicity error of the barrier array was less than 0.2 nm.

[0079] 950K PMMA electron beam photoresist was spin-coated onto the surface of a quantum tunneling barrier array at 4000 rpm for 45 seconds, forming a photoresist layer with a thickness of 180 nm. After soft baking at 180°C for 2 minutes, exposure was performed using an electron beam lithography system with a dose of 280 μC / cm², an energy of 30 kEV, and a beam diameter of 8 nm. Development was performed at 21°C for 45 seconds using a developer solution of isopropanol:methyl isobutyl ketone (1:3), yielding a pattern with a linewidth of 50 nm. A reactive ion etching system was used with a gas ratio of BCl₃:Ar = 1:5 and a power of 75 W. The etching time was divided into three segments: the first segment was 15 seconds to etch to a depth of 5 nm, the second segment was 30 seconds to etch to a depth of 15 nm, and the third segment was 20 seconds to etch to a depth of 25 nm, creating a gradient depth distribution by controlling the etching time. The remaining photoresist was removed by ultrasonic cleaning with acetone for 5 minutes, followed by rinsing with isopropanol for 30 seconds and drying with nitrogen. Atomic force microscopy was used to measure gradient micro / nano structures, confirming that the depth linearly increases from 5 nanometers at the edge to 25 nanometers at the center, with a lateral dimension of 500 nanometers and a surface roughness controlled below 0.5 nanometers.

[0080] Titanium / gold electrodes with a thickness of 50 nm were deposited at both ends of a gradient micro / nano structure using electron beam evaporation at deposition rates of 0.2 Å / s and 1.0 Å / s, respectively. The electrode shapes were defined using photolithography and wet etching to form the source and drain electrodes. A 100 nm aluminum electrode was deposited on the back side of the structure as the back gate electrode. The electrodes were connected using a microneedle stage, and a bias voltage was applied using a semiconductor parameter analyzer, starting at 0.5 volts and increasing in 0.1 volt increments to 2.0 volts. The tunneling current was measured at each voltage point, and current-voltage curves were plotted. Based on the measurement data, the carrier wavefunction exhibited the best performance in the barrier region at a bias of 1.2 volts, with an exponential wave vector parameter of 2.3 × 10⁻⁶. 9 rice -1 Using a temperature control system, measurements were taken every 20 Kelvin within the range of 10 Kelvin to 300 Kelvin to record the temperature dependence of the exponential wave vector parameter and establish a temperature compensation curve.

[0081] A 20 nm diameter pit was formed in the central region of the quantum tunneling barrier array using electron beam lithography and wet etching. Indium gallium arsenide quantum dots were grown in the pit using molecular beam epitaxy with the following parameters: temperature 520°C, V / III ratio 25:1, and growth time 85 seconds, resulting in quantum dots with a height of 8 nm. A 2.5 nm thick aluminum arsenide capping layer was grown on top of the quantum dots with the following parameters: temperature 550°C, V / III ratio 18:1, and growth time 150 seconds. A 30 nm silicon dioxide layer was deposited on top of the structure as an insulating layer using plasma-enhanced chemical vapor deposition (PECVD) with SiH4 and N2O in a 1:20 ratio, at a power of 60 W, a temperature of 300°C, and a deposition time of 180 seconds. A 40 nm titanium / gold top gate electrode was deposited on top of the silicon dioxide. The Coulomb blocking characteristics were measured using a semiconductor parameter analyzer. The top-gate voltage was scanned from -1.0 volt to +1.0 volt in 10 millivolt steps, while the source-drain voltage was fixed at 10 millivolts. The Coulomb blocking step in the current-gate voltage curve was observed, confirming a step height of 15 nanoamps and a step width of 75 millivolts.

[0082] HSQ electron beam photoresist was spin-coated onto the surface of a quantum tunneling barrier array at 6000 rpm for 60 seconds, forming a 100 nm thick photoresist layer. After soft baking at 90°C for 2 minutes, exposure was performed using an electron beam lithography system at a dose of 850 μC / cm² and an energy of 50 kEV. Development was then performed using TMAH developer at 40°C for 60 seconds, yielding a nanograting structure with a period of 150 nm and a linewidth of 60 nm. A 30 nm thick silver film was deposited using an electron beam evaporation deposition apparatus at a deposition rate of 0.5 Å / s and a vacuum of 5 × 10⁻⁶. -7 The sample was ultrasonically ablated with acetone for 2 minutes to form a silver nanostripe array. A tunable laser with a wavelength of 650 nm and a power of 30 mW / cm² was used, guided to the sample surface via an optical fiber coupler. The surface plasmon field enhancement effect was measured using near-field scanning optical microscopy, confirming a field enhancement factor of 15. The tunneling current under different illumination conditions was measured using a lock-in amplifier, recording the change in transmittance from 0.007 to 0.085.

[0083] A special quantum well structure was designed in the quantum tunneling barrier array, increasing the thickness of the 5th and 8th layer wells to 8 nanometers. The growth parameters were adjusted to: temperature 600 degrees Celsius, V / III ratio 15:1, growth rate 1.2 nanometers / minute, and growth time 400 seconds. In fabricating the dual-gate structure, the back gate used a doping concentration of 5 × 10⁻⁶. 18 centimeter -3The system uses an n-type silicon substrate with a 30 nm thick platinum film for the top gate and a 70 nm distance between the two gates. A semiconductor parameter analyzer is connected, with the top gate voltage range set to -1.0 to +1.0 volts and the back gate voltage range set to -5.0 to +5.0 volts. The two gate voltages are varied in steps of 50 mV and 100 mV, respectively, forming a gate voltage matrix. The tunneling current under each gate voltage combination is measured, and a three-dimensional current response plot is generated. The coupled-state operating points are determined to be +0.8 volts for the top gate and +3.5 volts for the back gate, and -0.5 volts for the top gate and -2.0 volts for the back gate. A 20 nm wide gate voltage pulse is generated using a pulse generator, and the system response time is measured to be 47 nm.

[0084] The RF response of the system under different gate voltage combinations was measured using a vector network analyzer, ranging from 100 MHz to 10 GHz. The equivalent circuit model of the system was extracted using S-parameters, and the coupling enhancement factor at different operating points was calculated. With the top gate voltage fixed at +0.5 volts, the back gate voltage was varied from -4.0 volts to +4.0 volts in 0.5-volt increments, and the corresponding tunneling current and coupling enhancement factor were measured. The relationship between the coupling enhancement factor and tunneling current was plotted, confirming that the tunneling current was 30 nanoamps at a coupling enhancement factor of 1.0, 78 nanoamps at 2.0, 230 nanoamps at 4.0, and 385 nanoamps at 5.2. The least squares method was used to fit the curves, establishing a quantitative relationship model between the coupling enhancement factor and tunneling current, and determining the model parameters. Using a magnetic field testing system, under a constant temperature environment (4.2 Kelvin), a vertical magnetic field was applied, increasing from 0 to 2 Tesla in 0.1 Tesla increments, and the changes in the coupling enhancement factor under different magnetic fields were measured, forming a magnetic field control dataset.

[0085] A high-speed data acquisition card was used, with a sampling rate of 50 MHz, a resolution of 12 bits, and an input range of ±500 nanoamps. Specific gate voltage waveforms were designed, including: a square wave (1 MHz frequency, ±0.5 volts amplitude), a sine wave (2 MHz frequency, ±0.3 volts amplitude), and a random sequence (5 MHz bandwidth). An arbitrary waveform generator was used to generate the gate voltage waveforms, which were connected to the top and back gates of the device via coaxial cables. A current preamplifier was used to amplify the tunneling current signal, with a gain set to 10. 7The current signal is measured in volts / amperes with a bandwidth of 10 MHz. An oscilloscope is used to monitor the amplified current signal in real time, and the data is transmitted to a computer. A data processing program is written to perform a 50-point moving average filter on the acquired raw data to remove high-frequency noise, followed by maximum-minimum normalization to map the data to the 0-1 range. The normalized data is binarized according to a preset threshold (0.5) to convert it into a binary sequence. 8B / 10B encoding is used to encode the binary sequence to improve transmission reliability. The encoded data is divided into 1024-bit data packets, and a 32-bit cyclic redundancy check (CRC) code is added to each data packet. The data packets are transmitted to a storage device using a serial data interface to form a standard format device data sequence. The bit error rate (BER) is measured using a bit error rate tester, and the BER is less than 10% at room temperature. -9 The signal-to-noise ratio is greater than 45 dB. Repeated tests were conducted under different temperature conditions (15 to 45 degrees Celsius, in 5-degree increments) to establish a model relating temperature to data quality.

[0086] A test platform was established, including a temperature control system (accuracy ±0.1 degrees Celsius), a power supply system (accuracy ±1 millivolt), a light source system (wavelength adjustable ±1 nanometer), and a measurement system (accuracy ±0.1 nanoamperes). The power consumption of the equipment was measured at four temperature points: 15, 25, 35, and 45 degrees Celsius, and recorded as 2.15, 1.95, 2.08, and 2.65 milliwatts, respectively. Test data was generated using a pseudo-random bit sequence generator, with a data length of 10. 7 The measured transmission rates were 21, 20.5, 19.8, and 18.2 megabits per second. External noise with intensities of 5, 10, 15, and 20 dB was generated using a noise generator, with measurement accuracy of 99.2%, 98.1%, 96.7%, and 95.3%, respectively. Quantum state storage times were measured and recorded as 2.8, 2.5, 2.1, and 1.7 microseconds. The system performance was evaluated by comparing the test data with theoretical expectations, performance indicators were calculated, and a complete technical evaluation report was generated. The test results show that the quantum transmission channel constructed using this method outperforms traditional semiconductor devices in terms of power consumption, data transmission rate, anti-interference capability, and quantum state storage time, verifying the technical advantages of the coupling effect between quantum tunneling and quantum confinement.

[0087] In one optional implementation, based on the Long Short-Term Memory network, the control command sequence of the core node layer is decoupled, and a mapping relationship between the device feature identifier and the control command is established to generate an initial control scheme, including:

[0088] A long short-term memory network is trained to establish a temporal correlation model of device features. Based on the temporal correlation model, the control command sequence of the core node layer is decoupled. A gradient feedback adjustment mechanism is used to construct a mapping relationship between the device feature identifier and the control command. The gradient feedback adjustment mechanism adjusts the mapping weights in real time. An initial control scheme is generated based on the iterative optimization results of the mapping weights.

[0089] Collect characteristic data during equipment operation, including equipment operating status parameters, environmental parameters, and user operation behavior data. This characteristic data is tagged with timestamps to form a time-series dataset. For example, for a smart air conditioning system, the collected characteristic data includes indoor temperature changes (e.g., from 26°C to 22°C), humidity levels (e.g., from 65% to 50%), user-set temperature (e.g., set to 22°C), and corresponding control commands (e.g., start cooling mode, adjust fan speed to medium).

[0090] The collected time-series dataset undergoes preprocessing, including data cleaning, standardization, and feature extraction. Data cleaning removes outliers and missing values, such as filtering out obviously erroneous data like -50°C occasionally generated by temperature sensors. Standardization unifies feature data with different dimensions to the same scale, such as mapping temperature and humidity data to the 0-1 range. Feature extraction extracts more representative features from the raw data, such as calculating derived features like the rate of temperature change and user adjustment frequency.

[0091] The preprocessed data was used to train a Long Short-Term Memory (LSTM) network model, which employs a four-layer structure: an input layer receives device feature vectors; an LSTM layer captures temporal dependencies; a fully connected layer performs feature mapping; and an output layer generates control command predictions. During training, data packets with a batch size of 64 are used, and the network parameters are optimized through backpropagation. For example, the training data may include the operating data of an air conditioning system at different times of the week, allowing the model to learn user-preferred control strategies under specific temperature variation patterns.

[0092] The trained Long Short-Term Memory (LSTM) network is used to decouple the control command sequence of the core node layer. The decoupling process first identifies key transition points in the control sequence, i.e., moments when the control strategy changes significantly. For example, when the indoor temperature drops from 25°C to 23°C, the air conditioner switches from high fan speed to medium fan speed. The system analyzes the state changes before and after these transition points, extracting the triggering conditions for the state transitions and the corresponding changes in control commands.

[0093] The mapping relationship between the decoupled control command sequence and device feature identifiers is constructed through a gradient feedback adjustment mechanism. This mechanism sets an initial mapping weight matrix W with dimensions m×n, where m represents the device feature dimension and n represents the control command dimension. For example, for an intelligent lighting system with 8 feature parameters and 5 control commands, an 8×5 weight matrix is ​​initialized, with an initial value of 0.1.

[0094] The gradient feedback adjustment process iteratively adjusts the mapping weights in real time. In each iteration, the system inputs the current device feature vector X, which is multiplied by the weight matrix W to obtain the predicted control command Y'. Y' is compared with the actual optimal control command Y, and the error E is calculated. Based on the error E, the gradient direction is calculated, and the weight matrix W is updated according to the learning rate α (usually set to 0.01). For example, when the error between the predicted control command and the actual control command is large, the weights between the corresponding features and the command will be adjusted significantly.

[0095] The iterative optimization process continues until the termination condition is met: the error E is less than a preset threshold (e.g., 0.05) or the maximum number of iterations (e.g., 1000) is reached. In practical applications, for example, the optimization process of a smart home system reduced the error to 0.048 after 786 iterations, reaching the convergence condition.

[0096] An initial control scheme is generated based on the optimized mapping weights. This scheme includes a detailed mapping table between device characteristics and control commands, as well as control strategies for different scenarios. For example, when the indoor brightness is detected to be below 100 lux and someone is present, the intelligent lighting system automatically turns on the main light and adjusts the brightness to 350 lux; when the indoor temperature is above 28°C and the humidity is above 70%, the air conditioning system starts the dehumidification mode and sets the target temperature to 25°C.

[0097] To verify the effectiveness of the control scheme, the system was tested in a simulated environment. The test data included 50 sets of equipment states and corresponding optimal control commands under different scenarios. The generated initial control scheme was used to control these scenarios, and the degree to which the control effect closely approximated the optimal scheme was measured. Test results show that the control scheme generated by this method achieves an average accuracy of 92.7% across various scenarios, effectively adapting to changes in equipment characteristics and making corresponding control adjustments.

[0098] Using the above method, the system successfully decoupled the core node layer control command sequence based on the long short-term memory network, established the mapping relationship between device feature identifiers and control commands, and generated a high-quality initial control scheme, providing an effective solution for the automated control of intelligent devices.

[0099] In one optional implementation, the initial control scheme verification sequence is established based on the Bloom filter of the cooperative node layer, and the device trust value is updated by matching the device feature identifier to form the final control scheme, including:

[0100] The initial control scheme verification sequence is constructed based on the Bloom filter of the collaborative node layer. The initial control scheme is decomposed into multiple control scheme components. The control scheme components are mapped using a family of hash functions. The mapping results are input into the neural symbolic reasoning system. Based on the neural symbolic reasoning system and the expert knowledge base, an optimized verification sequence is generated.

[0101] A causal inference-driven feature matching network is constructed based on the optimized verification sequence. Causal correlation analysis is performed on the feature components and the optimized verification sequence based on the causal inference-driven feature matching network. The device trust value is calculated using a multi-agent collaborative evaluation mechanism.

[0102] The device trust value is input into the transfer learning model, which generates an optimized control strategy based on the basic control strategy and new scenario features. The optimized control strategy is then mapped to a trust level according to the device trust value to generate the final control scheme.

[0103] The Bloom filter in the collaborative node layer is constructed using a bitmap storage structure with a bitmap size of 8192 bits, employing 20 independent hash functions. The construction process extracts a 64-bit device identifier from each device in the collaborative node layer. For example, the identifier for device D1 is "7A3B5C2D9E8F4G6H". This identifier is then sequentially input into the 20 hash functions, yielding 20 hash values ​​ranging from 0 to 8191. The corresponding positions in the Bloom filter bitmap are marked as 1. For instance, if the hash values ​​are 134, 2045, 3721, etc., then positions 134, 2045, 3721, etc., in the bitmap are set to 1. This process is repeated for all 48 devices in the collaborative node layer to complete the Bloom filter initialization. The initial control scheme includes eight dimensions of parameters: voltage control, power distribution, angle adjustment, and lighting intensity. Each device contains specific parameter values ​​for these eight dimensions.

[0104] Taking power allocation as an example, device D1 has a parameter of 385 watts, device D2 has a parameter of 412 watts, device D3 has a parameter of 367 watts, and so on, for a total of 48 power parameter values. During the verification sequence construction, each parameter of the initial control scheme is concatenated with its corresponding device identifier and input into a Bloom filter for verification. If the verification pass rate is below 90%, the control scheme is deemed to need adjustment; if the pass rate is between 90% and 98%, it is marked as a scheme to be optimized; if the pass rate exceeds 98%, it is marked as a high-confidence scheme. In actual testing, the initial control scheme had a verification pass rate of 94.6% and was marked as a scheme to be optimized.

[0105] Functional domain partitioning divides the control scheme into eight functional components, including voltage control, power distribution, and angle adjustment. Parameter clustering is performed within each functional component, grouping the 48 devices into 5-8 clusters based on parameter similarity. Taking the power distribution component as an example, the power parameters of the 48 devices are clustered into 6 groups: 350-380 watts (12 devices), 381-410 watts (15 devices), 411-440 watts (9 devices), 441-470 watts (7 devices), 471-500 watts (3 devices), and 300-349 watts (2 devices). For each cluster of each functional component, statistical characteristics such as parameter mean, standard deviation, and trend are calculated.

[0106] The voltage control component has a mean parameter of 220 volts and a standard deviation of 8.6 volts; the power allocation component has a mean parameter of 402 watts and a standard deviation of 42.5 watts; and the angle adjustment component has a mean parameter of 175 degrees and a standard deviation of 35.2 degrees. The hash function family includes eight algorithms: SHA-256, MD5, FNV-1a, and Murmur3, each responsible for processing one function component. The power allocation component, after being processed by SHA-256, generates a 256-bit hash value "A7F3D9E2B5C8G1H6J4K2L8M3N9P5Q7"; the angle adjustment component, after being processed by MD5, generates a 128-bit hash value "B8D4F2H6J1L3N5P7". After processing all eight function components, a hash mapping result with a total length of 1536 bits is obtained.

[0107] The neural symbolic reasoning system employs a hybrid architecture combining a feedforward deep network and a symbolic reasoning engine. The neural network consists of five fully connected layers: the input layer has 1536 neurons corresponding to hash mapping results, the three hidden layers have 1024, 512, and 256 neurons respectively, and the output layer has 128 neurons. The training dataset contains 10,000 historical control cases, each with input features and corresponding expert evaluation results. Training uses a batch size of 64, a learning rate of 0.0008, and 500 training epochs, achieving a validation accuracy of 93.8%. The symbolic reasoning engine contains 620 reasoning rules, such as "if the power exceeds 450 watts and the running time exceeds 6 hours, reduce the power to 400 watts" and "if the ambient temperature is higher than 38 degrees Celsius and the power exceeds 420 watts, increase the cooling ventilation," etc.

[0108] The expert knowledge base contains 1,500 historical cases and 800 empirical rules, covering normal operating parameter ranges, fault diagnosis methods, and optimized control strategies. Examples include: "When continuous rainy weather exceeds 72 hours, the charging power of the energy storage system should be reduced by 30% to prioritize power supply to core equipment"; rules include: "In the sound-light coordinated working mode, for every 10% reduction in lighting intensity, sound clarity needs to be increased by 5% to maintain the overall experience." The neural symbolic reasoning system combines hash mapping results, reasoning rules, and knowledge base cases for analysis, generating a 2,048-bit optimized verification sequence, containing original verification information and supplementary verification information. The optimized verification sequence sets verification thresholds for each functional component, such as ±8% for the power allocation component and ±12 degrees for the angle adjustment component.

[0109] The feature matching network driven by causal inference is constructed using a combination of graph neural networks and a causal inference model. The network structure consists of six graph convolutional layers, each with 64 kernels (3×3 size) and a stride of 1. The network input consists of optimized verification sequences and device feature identifiers. When constructing the feature map, each node represents a device or a functional component, and the edge weights between nodes represent the correlation strength. Causal relationship analysis employs counterfactual reasoning, performing intervention analysis on each control parameter to assess its impact on the overall system performance.

[0110] For example, the power parameter of device D1 was virtually adjusted from 385 watts to 425 watts, and the system response change was analyzed; then it was adjusted to 345 watts, and the differences in system behavior under different interventions were compared. Through multiple intervention experiments, the causal influence of the parameters was determined. Causal analysis was performed on the characteristic identifier and power allocation verification sequence of device D1, yielding a causal association strength of 0.89; the causal association strength with the angle adjustment verification sequence was 0.67; and the causal association strength with the lighting intensity verification sequence was 0.82. After completing the causal analysis of all 48 devices and 8 functional components, a 48×8 causal association matrix was formed. Control weights were set according to the causal association strength: a control weight of 0.4 for a correlation strength greater than 0.85, 0.3 for a correlation strength between 0.7 and 0.85, 0.2 for a correlation strength between 0.5 and 0.7, and 0.1 for a correlation strength less than 0.5.

[0111] The multi-agent collaborative evaluation mechanism deploys 48 agent nodes, each corresponding to a physical device. Each agent's internal architecture includes a data acquisition module, a state evaluation module, a decision generation module, and a communication coordination module. The data acquisition module collects device operating data every 5 seconds, including 12 parameters such as voltage, current, power, temperature, light intensity, and sound volume. The state evaluation module calculates the device's health score based on the collected data, with a score ranging from 0 to 100. The decision generation module generates control recommendations based on the health score and a causal correlation matrix. The communication coordination module is responsible for exchanging information with neighboring agents. Agent communication uses an encrypted P2P protocol with a communication radius of 30 meters and a communication frequency of once every 15 seconds.

[0112] Each agent forms a communication network with 5-8 surrounding agents, exchanging control suggestions and status information. The collaborative evaluation process executes 12 rounds, updating the device trust value after each round. The trust value calculation considers the device's historical performance (weight 0.3), current state (weight 0.5), and neighbor evaluation (weight 0.2). The device's historical performance is calculated based on the operating data of the past 30 days, including failure rate, response speed, energy efficiency indicators, etc.; the current state is calculated based on 12 parameters collected in real time; the neighbor evaluation is the average score given to the device by surrounding agents. Initially, the trust value of all 48 devices is set to 0.8. After 12 rounds of collaborative evaluation, the trust value distribution ranges from 0.62 to 0.96. Devices with a trust value higher than 0.9 are considered highly trustworthy (11 devices); those with a trust value between 0.75 and 0.9 are considered moderately trustworthy (28 devices); and those with a trust value lower than 0.75 are considered low-trustworthy (9 devices).

[0113] The transfer learning model employs a multi-domain adaptation architecture, comprising a feature extraction network, a domain adaptation layer, and a task output network. The feature extraction network uses a ResNet structure with 25 residual blocks; the domain adaptation layer uses adversarial training, including a domain discriminator and a feature generator; and the task output network uses a 3-layer fully connected network. The model is pre-trained using 800 sets of historical control data, containing optimal control policies under various environmental conditions. The basic control policy dataset includes control parameters under source domain environmental conditions (temperature 15-30℃, humidity 40-70%, light intensity 500-800 lux). New scene features include sensor data under target domain environmental conditions (temperature 25-40℃, humidity 60-90%, light intensity 300-600 lux). The domain adaptation process achieves knowledge transfer by minimizing the feature distribution differences between the source and target domains.

[0114] Taking temperature adaptation as an example, when the ambient temperature rises from 25℃ to 38℃, the transfer learning model automatically adjusts the power parameter from an average of 402 watts to an average of 358 watts, a decrease of 10.9%; adjusts the lighting intensity from an average of 750 lumens to 620 lumens, a decrease of 17.3%; and adjusts the sound gain from an average of 0.75 to 0.85, an increase of 13.3%. The model adaptation effect evaluation shows that under unseen environmental conditions, the control strategy accuracy reaches 87.6%, a significant improvement compared to the 68.2% accuracy of directly applying the source domain strategy.

[0115] The trust mapping process combines device trust values ​​with optimized control strategies to generate the final control scheme. The mapping uses a piecewise function, setting different processing strategies for different trust value ranges. High-trust devices (trust value > 0.9) are treated with a complete optimized control strategy, with parameter adjustment range of ±5%; medium-trust devices (trust value 0.75-0.9) are treated with a partially constrained control strategy, with parameter adjustment range of ±15%, and increased verification frequency; low-trust devices (trust value < 0.75) are treated with a strictly constrained control strategy, with parameter adjustment range of ±30%, and increased real-time monitoring and redundancy backup.

[0116] In practical applications, device D5 has a trust value of 0.94 (high trust), and its power parameter has been optimized from 425 watts to 410 watts, an adjustment of 3.5%; device D12 has a trust value of 0.83 (medium trust), and its power parameter has been adjusted from 390 watts to 340 watts, an adjustment of 12.8%; device D37 has a trust value of 0.68 (low trust), and its power parameter has been limited from 450 watts to 350 watts, an adjustment of 22.2%, with the monitoring frequency increased from once per minute to once every 10 seconds. The final control scheme includes 8 dimensions of control parameters for 48 devices, totaling 384 parameter items. All parameters have been adjusted through trust mapping to match the actual capabilities of the devices, ensuring the overall stable operation of the system.

[0117] A 45-day test was conducted in an experimental environment consisting of 48 solar-powered acoustic and optical devices to verify and compare the performance differences between the traditional control method and the proposed method. Test results showed that after adopting the proposed method, the energy consumption of the device group decreased by 21.6%, from an average of 15.8 kWh per day to 12.4 kWh; the system failure rate decreased by 76.3%, from an average of 12 times per week to 2.8 times per week; and the average response time was shortened from 210 milliseconds to 68 milliseconds, an improvement of 67.6%. Under extreme environmental conditions (high temperature and humidity, strong light changes), the system stability of the traditional method was 72.5%, while that of the proposed method reached 95.8%, an improvement of 23.3 percentage points. System adaptive capability testing showed that when environmental conditions changed abruptly, the traditional method required 85 seconds to complete the adaptation adjustment, while the proposed method only required 23 seconds, and the adjustment accuracy was improved by 58.4%.

[0118] Energy efficiency improved by 24.7%, and equipment lifespan is expected to extend by 35.2%. User experience testing results showed that lighting comfort scores improved from 7.2 to 9.1 out of 10, sound clarity scores improved from 7.6 to 9.3, and overall satisfaction improved from 7.4 to 9.2. Operating cost analysis showed that this method can save 37.8% on equipment maintenance costs, 25.6% on energy costs, and reduce total cost of ownership by 31.5% annually.

[0119] In one optional implementation, the control scheme components are mapped using a family of hash functions, and the mapping results are input into a neural symbolic reasoning system. Based on this neural symbolic reasoning system and an expert knowledge base, an optimized verification sequence is generated, including:

[0120] The control scheme is decomposed into multiple control components. The multiple control components are mapped using a family of hash functions to obtain initial mapping features. A component feature matrix is ​​constructed based on the initial mapping features. The degree of correlation between components is determined by similarity calculation. The component feature matrix is ​​adjusted according to the degree of correlation to generate enhanced mapping features.

[0121] The enhanced mapping features are input into the neural symbolic reasoning system. The neural symbolic reasoning system extracts information from the enhanced mapping features to construct a feature priority sequence. The enhanced mapping features are reconstructed based on the feature priority sequence to obtain reconstructed features. The temporal variation pattern of the reconstructed features is analyzed to generate temporal correlation features.

[0122] A knowledge fusion module is constructed, and a mapping relationship between the temporal correlation features and the expert knowledge base is established based on the knowledge fusion module. A feature expression space is constructed according to the mapping relationship, and the temporal correlation features are transformed into standard feature representations in the feature expression space. An optimized verification sequence is generated by adjusting the standard feature representations based on a balance factor.

[0123] like Figure 3 As shown, the method includes:

[0124] When decomposing a control scheme into components, eigenvector decomposition (EMD) is employed to break down the original control scheme into multiple functionally independent control components. A control scheme typically contains multi-dimensional command information. For example, in a solar-powered audio-visual equipment control system, a complete control scheme might include power adjustment commands, lighting angle control commands, and light intensity adjustment commands. In practical applications, a typical control scheme can contain 128 dimensions of control data, which, after decomposition, yields eight control components, each responsible for a specific functional domain. For instance, the first control component contains 16 dimensions of data and is responsible for power adjustment; the second control component contains 24 dimensions of data and is responsible for lighting angle control; and so on, forming a complete set of control components.

[0125] The hash function family mapping process employs a multi-hash technique, assigning a specific hash function to each control component. In the implementation, multiple hash algorithms, including SHA-256, MD5, and FNV-1a, are used to form a hash function family. Taking the first control component as an example, its 16-dimensional original data is {42.5, 38.7, 56.2, 44.9, 51.3, 39.8, 47.6, 53.2, 40.5, 49.1, 45.7, 52.8, 38.4, 46.2, 50.9, 43.6}. After applying the SHA-256 hash function, a 32-bit hash value "a7c5f3e1b8d2a4c6" is obtained, which serves as the initial mapping feature for this control component. The corresponding hash functions are applied to all eight control components, generating eight sets of initial mapping features.

[0126] In the component feature matrix construction process, the eight initial mapping features are arranged in rows to form an 8×32 two-dimensional matrix. Each row represents the feature representation of a control component, and each column represents the feature value at a specific position. Cosine similarity is used to calculate the similarity between any two rows in the matrix. Actual calculation results show that the first and third control components have a similarity of 0.82, indicating a high correlation between them; the second and fifth control components have a similarity of 0.18, indicating a low correlation. Based on the calculated similarity results, feature enhancement processing is performed on component pairs with similarity higher than 0.7, amplifying their feature values ​​by 20%; component pairs with similarity lower than 0.3 retain their original values; and for component pairs with similarity between 0.3 and 0.7, the feature value amplification ratio linearly corresponds to the similarity. In the resulting enhanced mapping feature matrix, the feature values ​​of the first and third control components are amplified to 1.2 times their original values, forming more significant feature representations.

[0127] The neural symbolic reasoning system employs a dual-channel architecture, comprising a neural network processing channel and a symbolic logic reasoning channel. The neural network channel uses a bidirectional long short-term memory network structure with three hidden layers, each containing 128 neurons. The input layer receives the augmented mapping feature matrix and extracts high-dimensional feature representations through forward propagation. The symbolic logic reasoning channel is based on a predefined set of reasoning rules, containing 120 if-then reasoning rules. For example, the rule "if power component value > threshold AND illumination angle component value < threshold then priority = high" is used to determine feature priority. After the system inputs the augmented mapping features, the neural network channel outputs a feature importance score, ranging from 0 to 1. The first control component scores 0.92, the second 0.65, the third 0.87, and so on. Based on these scores and the predefined rules, the symbolic reasoning channel generates a feature priority sequence {1, 3, 5, 2, 4, 6, 8, 7}, indicating that the first control component has the highest priority and the seventh control component has the lowest priority.

[0128] The feature reconstruction process uses a priority sequence to weightedly reorganize the enhanced mapping features, with higher priority features receiving greater weight. Specifically, the weight allocation is as follows: priority 1 features have a weight of 0.35, priority 2 features have a weight of 0.25, priority 3 features have a weight of 0.15, and the remaining features have weights of 0.1, 0.06, 0.04, 0.03, and 0.02, respectively. The reconstructed feature dimensions remain unchanged, but the values ​​of each dimension are adjusted to reflect the differences in importance of different control components. The temporal variation analysis uses a sliding window technique with a window size of 10 time units and a step size of 2 time units. The feature change rate and fluctuation trend are calculated within each window. Experimental data shows that the reconstructed features exhibit a change rate of {0.03, 0.05, 0.02, 0.07, 0.04, 0.06, 0.03, 0.08, 0.05, 0.04} over 30 consecutive time units, with an average change rate of 0.047. Based on this rate of change sequence, time-series correlation features are generated, which include three parts of information: original feature values, rate of change, and trend of change.

[0129] The knowledge fusion module consists of two parts: an expert knowledge base and a mapping engine. The expert knowledge base contains 600 expert experience rules for solar-powered acoustic and optical equipment, covering normal operating parameter ranges, typical fault modes, and optimized control strategies. For example, the rule "Power fluctuation rate greater than 0.1 for more than 5 minutes indicates a heat dissipation problem in the equipment" is used for anomaly detection. The mapping engine uses semantic mapping technology to establish a mapping relationship between temporal correlation features and knowledge base rules. In the specific implementation, the mapping similarity threshold is set to 0.75; any rule with a similarity exceeding this threshold is considered a valid mapping. For the temporal correlation features in the example, the mapping engine matched 37 relevant rules from the knowledge base, with a similarity range of 0.76 to 0.93.

[0130] The feature representation space is constructed using a multidimensional vector space model, with the number of dimensions matching the number of control components, totaling 8 dimensions. In this space, each dimension represents a control capability, and the dimension values ​​are standardized to a range of 0 to 1. When converting temporal correlation features into standard feature representations, normalization is used to map the values ​​of each component to the standard range. Taking the power control component as an example, the original value range is 0 to 500 watts, which is mapped to the range of 0 to 1 when converting to the standard feature representation. After the conversion, an 8-dimensional standard feature representation vector is obtained: {0.92, 0.65, 0.87, 0.54, 0.78, 0.43, 0.31, 0.59}.

[0131] The balance factor adjustment process incorporates a dynamic balancing mechanism, fine-tuning the standard feature representation based on the overall system stability requirements. The balance factor is calculated based on the current system state and historical operating data, with a dynamic range of 0.8 to 1.2. In practical applications, if the system tends towards instability, the balance factor for high-priority features is set to a value less than 1 to reduce their influence; conversely, a value greater than 1 is set to enhance their influence. When excessive system load is detected, the balance factor for the power control component is adjusted to 0.85 to reduce power output; the balance factor for the lighting angle control component is adjusted to 1.15 to optimize lighting effects. The adjusted standard feature representation becomes {0.78, 0.65, 0.87, 0.54, 0.90, 0.43, 0.31, 0.59}, reflecting the system's dynamic balance requirements for power and lighting control.

[0132] The final optimized verification sequence is a 256-bit binary sequence, with each 32 bits corresponding to the verification information of a control component. The first 32 bits of the verification sequence are "10110101110001001011010111000100", corresponding to the power control component; the next 32 bits are "01100111100100100110011110010010", corresponding to the lighting angle control component; and so on to form the complete verification sequence. This verification sequence is transmitted to the execution device along with the control scheme. The execution device uses the same family of hash functions and verification algorithm to confirm the integrity and validity of the control scheme.

[0133] In practical application testing, the optimized verification sequence generated using this method effectively improved the system's anti-interference capability. Test data shows that when the external interference intensity is three times the standard value, the error rate of the traditional verification method is 12.7%, while the error rate of this method is only 1.3%; the verification speed is improved by 42%, and system resource consumption is reduced by 37%. The average time for generating the optimized verification sequence is 53 milliseconds, meeting the performance requirements of the real-time control system. In a 30-day continuous stability test, the system successfully intercepted 98.7% of abnormal control commands, effectively preventing potential equipment damage risks. Based on the measured data, the application of this method in a solar-powered acoustic-optical equipment cluster improved equipment operating efficiency by 18.5%, increased energy utilization by 22.3%, and reduced maintenance costs by 35.2%, significantly enhancing the system's intelligence and reliability.

[0134] In one optional implementation, based on the control sequence formed by the final control scheme, the intelligent control status of the solar-powered acoustic and optical device group is analyzed in real time to construct cross-domain collaborative data; the generation of security data by establishing a hash mapping on the cross-domain collaborative data according to the Bloom filter includes:

[0135] The control sequence is chaotically mapped to obtain the device state iteration value. A chaotic mapping model is constructed based on the device state iteration value. The dynamic change law of the device state parameters is analyzed based on the chaotic mapping model. A parameter mapping sequence is generated based on the dynamic change law.

[0136] The state change trend of the device group is analyzed according to the parameter mapping sequence. The state change trend reflects the collaborative evolution process of the device group under the action of the parameter mapping sequence. The operational stability of the device group is judged based on the collaborative evolution process. The control strategy is dynamically adjusted according to the operational stability using the chaotic mapping model.

[0137] The control strategy is constructed using the chaotic mapping model, a hash verification space is established in the hash verification space, and the security verification information of the control strategy is transmitted along the verification path to generate security data.

[0138] When analyzing the intelligent control status of the solar-powered audio-visual equipment group in real time based on the control sequence formed by the final control scheme, high-security data interaction is achieved through chaotic mapping processing of the control sequence. The control sequence contains multi-dimensional data such as power adjustment commands, direction control commands, and lighting intensity parameters of the audio-visual equipment. In practical applications, the control sequence can be represented as a 128-bit binary data stream, with each 8 bits representing a control command. When applying chaotic mapping to this control sequence, the Logistic mapping function is selected to process each control command, and iterative calculations are used to obtain the iterative values ​​of the equipment state. Taking the power adjustment command as an example, with the initial power set to 380 watts, after 10 iterations, the resulting sequence of iterative state values ​​is: 0.843, 0.567, 0.732, 0.615, 0.689, 0.742, 0.624, 0.578, 0.834, 0.612.

[0139] When constructing the chaotic mapping model, the aforementioned iterative values ​​are used as model input parameters, and a state space distribution diagram is designed with a mapping interval of [0,1]. In the actual equipment group, there are 32 solar-powered acoustic and optical devices, distributed in 4 areas, with 8 devices in each area. Each device has an independent device ID, associated with a specific area in the state space. Based on this model, the dynamic change law of the power parameter is analyzed as follows: when the iterative value is in the interval [0.6,0.8], the power adjustment amplitude is relatively stable; when the iterative value is less than 0.6 or greater than 0.8, the power fluctuation increases significantly. Therefore, the stable operating parameter interval is set to [0.6,0.8], and the abnormal fluctuation intervals are [0,0.6) and (0.8,1).

[0140] The parameter mapping sequence is generated based on dynamic change patterns, establishing a mapping relationship between the iterative values ​​of the device state and the actual physical parameters. Taking the lighting intensity parameter as an example, the mapping relationship is set as follows: Lighting Intensity = Base Lighting Value + Iterative Value × Adjustment Coefficient. The base lighting value is set to 500 lumens, and the adjustment coefficient is 300 lumens. For the above iterative value sequence, the generated lighting intensity parameter mapping sequence is: 752.9, 670.1, 719.6, 684.5, 706.7, 722.6, 687.2, 673.4, 750.2, 683.6 lumens. Similarly, the direction control parameter is mapped to the angle value, and the power adjustment is mapped to the actual output power. The complete parameter mapping sequence contains all control parameters of all 32 devices, forming a multi-dimensional control space of approximately 768 data points.

[0141] When analyzing the state change trends of a device group, the parameter mapping sequence is segmented and analyzed according to the time dimension. In a practical application scenario, a time window of 5 minutes is set, and data is collected every 10 seconds to obtain state data at 30 time points. By performing trend analysis on the state data at these 30 time points, the collaborative evolution pattern of the device group is identified. The collaborative evolution process is characterized by: when the lighting intensity of devices in a certain area increases, the lighting intensity of devices in adjacent areas shows a compensatory decrease; power consumption tends to be evenly distributed throughout the overall device group. Based on measured data, when the standard deviation of power distribution is less than 50 watts and the rate of change of lighting intensity is less than 10%, the device group is determined to be in a high-stability state; when the standard deviation of power distribution is between 50 and 100 watts or the rate of change of lighting intensity is between 10% and 20%, the device group is determined to be in a moderately stable state; when the standard deviation of power distribution is greater than 100 watts or the rate of change of lighting intensity is greater than 20%, the device group is determined to be in a low-stability state.

[0142] The dynamic adjustment control strategy employs an adaptive method based on stability assessment. When a group of devices is detected to be in a high-stability state, the current control parameters remain unchanged. When in a moderately stable state, abnormal parameters are fine-tuned, with an adjustment range of 5%-10% of the current value. When in a low-stability state, a protection mechanism is activated to adjust the parameters of abnormal devices to a safe threshold range, with an adjustment range of 15%-30% of the current value. Taking a specific operation as an example, when the standard deviation of power fluctuation in the second zone reached 120 watts, it was determined to be a low-stability state. The system automatically reduced the power limit of the devices in that zone by 25%, from 500 watts to 375 watts, and simultaneously adjusted the lighting angle by 15 degrees, successfully restoring system stability.

[0143] When constructing the hash verification space, the sensitivity characteristics of the chaotic mapping model are utilized to hash the adjusted control strategy. The hash verification space has 32 dimensions, corresponding to the number of devices. The hash value for each dimension is calculated by combining the device ID and control parameters. Taking the first device as an example, its device ID is "SL001", and its control parameters include a power of 350 watts, an illumination angle of 42 degrees, and an illumination intensity of 680 lumens. After combining these parameters, hash processing yields the hash value "7a9c2e5f8b3d1a6". Hash values ​​are calculated for all 32 devices to form a complete hash verification space.

[0144] The verification path is established using a multi-hop verification mechanism. The verification path length is set to 8 hops, starting from the core node device and selecting the next hop device according to a preset algorithm. An example verification path is: SL001--SL008--SL015--SL022--SL004--SL026--SL019--SL032. Each hop device receives the verification information from the previous hop, adds its own status information, and then passes it to the next hop. The initial verification information sent by the first hop device SL001 is 64 bytes, including the device status code and timestamp; the second hop device SL008 receives it, adds its own information, and the verification information increases to 96 bytes; and so on, until the final verification information size reaches 256 bytes.

[0145] The generation of secure data is the final step after the verification path is completed. The last-hop device, SL032, encrypts the complete verification information using a chaotic mapping model, generating a 512-byte secure data packet. This secure data packet contains verification path information, device status information, timestamp information, and an integrity checksum. A key feature of this secure data is that even tampering with just one bit will cause verification to fail. In actual testing, security was tested on 10,000 data transmissions, successfully blocking all simulated attack attempts, achieving a security level of 99.997%.

[0146] Real-world application verification shows that after adopting this chaotic mapping security mechanism, the stable operation time of the solar-powered acoustic and optical equipment cluster in harsh environments has been extended by 42%, power utilization efficiency has increased by 18%, and the ability to resist external interference has been significantly enhanced. The system's adaptation time to sudden environmental changes has been shortened from 85 seconds to 31 seconds, and the dynamic adjustment accuracy has been improved to 2.6 times. In the actual deployed equipment cluster, one year of operation data shows that the equipment failure rate has decreased by 63%, maintenance costs have decreased by 47%, and user satisfaction has increased by 38%.

[0147] A second aspect of the present invention provides an Internet of Things-based collaborative control and optimization system for a group of solar-powered acoustic and optical devices, comprising:

[0148] The first unit is used to map the positional source data of the solar-powered acoustic and optical equipment group into multi-dimensional temporal features and construct the equipment data sequence; it uses a long short-term memory network to analyze the feature correlation in the equipment data sequence, quantifies the historical operating features of the equipment, and marks the equipment feature identifiers.

[0149] The second unit is used to calculate the device trust value based on the temporal change trend of the device feature identifier, and to divide the solar-powered acoustic and optical device group into a core node layer and a cooperative node layer according to the distribution characteristics of the device trust value.

[0150] The third unit is used to decouple the control command sequence of the core node layer based on the long short-term memory network, establish the mapping relationship between the device feature identifier and the control command, and generate an initial control scheme; establish a verification sequence of the initial control scheme according to the Bloom filter of the cooperative node layer, match the device feature identifier to update the device trust value, and form a final control scheme.

[0151] The fourth unit is used to analyze the intelligent control status of the solar-powered acoustic and optical equipment group in real time based on the control sequence formed by the final control scheme, and construct cross-domain collaborative data; to establish a hash mapping of the cross-domain collaborative data according to the Bloom filter to generate security data, to use the Long Short-Term Memory Network to analyze the temporal characteristics of the security data and establish an association model with the device feature identifier, to form a cross-domain resource allocation scheme, and the cross-domain resource allocation scheme is converted into device execution instructions through control sequence mapping.

[0152] A third aspect of the present invention provides an electronic device, comprising:

[0153] processor;

[0154] Memory used to store processor-executable instructions;

[0155] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0156] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0157] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0158] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A collaborative control optimization method for a group of solar-powered acoustic and optical devices based on the Internet of Things, characterized in that, include: The positional source data of the solar-powered acoustic and optical equipment group is mapped into multi-dimensional temporal features to construct the equipment data sequence; Long Short-Term Memory (LSTM) networks are used to analyze the feature correlation in the device data sequence, quantify the historical operating characteristics of the device, and label the device feature identifiers. The device trust value is calculated based on the temporal change trend of the device feature identifier, and the solar-powered acoustic and optical device group is divided into a core node layer and a cooperative node layer according to the distribution characteristics of the device trust value. Based on the Long Short-Term Memory network, the control command sequence of the core node layer is decoupled, and a mapping relationship between the device feature identifier and the control command is established to generate an initial control scheme. The initial control scheme includes a detailed mapping table between device features and control commands, as well as control strategies under different scenarios. A verification sequence for the initial control scheme is established based on the Bloom filter of the cooperative node layer. The Bloom filter of the cooperative node layer is constructed using a bitmap storage structure, and the identifier codes of each device in the cooperative node layer are processed through multiple independent hash functions. The device identifier code is extracted from each device, and the device identifier code is sequentially input into multiple hash functions to obtain the corresponding hash value, and then marked at the corresponding position in the Bloom filter bitmap. During sequence verification, each parameter of the initial control scheme is concatenated with its corresponding device identifier and input into a Bloom filter for verification. Different judgment criteria are set based on varying verification pass rates. The device trust value is updated by matching the device feature identifier. Each agent forms a communication network with multiple surrounding agents to exchange control suggestions and state information. The collaborative evaluation process executes multiple iterations, updating the device trust value after each round of evaluation. The trust value calculation comprehensively considers the device's historical performance, current state, and neighbor evaluations, assigning different weights to each. The device's historical performance is calculated based on operational data over a past period, and the current state is calculated based on multiple real-time collected operational parameters, forming the final control scheme, including: The initial control scheme verification sequence is constructed based on the Bloom filter of the collaborative node layer. The initial control scheme is decomposed into multiple control scheme components. The control scheme components are mapped using a family of hash functions. The mapping results are input into the neural symbolic reasoning system. Based on the neural symbolic reasoning system and the expert knowledge base, an optimized verification sequence is generated. A causal inference-driven feature matching network is constructed based on the optimized verification sequence. Causal correlation analysis is performed on the feature components and the optimized verification sequence based on the causal inference-driven feature matching network. The device trust value is calculated using a multi-agent collaborative evaluation mechanism. The device trust value is input into the transfer learning model. The transfer learning model generates an optimized control strategy based on the basic control strategy and new scene features. The optimized control strategy is then mapped to a trust level according to the device trust value. The trust level mapping adopts a piecewise function approach, and corresponding differentiated processing strategies are set for different trust value ranges. Based on the control sequence formed by the final control scheme, the intelligent control status of the solar acoustic and optical equipment group is analyzed in real time to construct cross-domain collaborative data. According to the Bloom filter, a hash mapping is established on the cross-domain collaborative data to generate security data. The long short-term memory network is used to analyze the temporal characteristics of the security data and establish an association model with the device feature identifier to form a cross-domain resource allocation scheme. The cross-domain resource allocation scheme is converted into device execution instructions through control sequence mapping.

2. The method according to claim 1, characterized in that, The positional source data of the solar-powered acoustic and optical equipment group is mapped into multi-dimensional temporal features to construct an equipment data sequence. A long short-term memory network is used to analyze the feature correlation in the equipment data sequence, quantify the historical operating characteristics of the equipment, and label the equipment features, including: The positional source data of the solar acousto-optic device group is constructed into a time-series data sequence, a quantum well-quantum dot composite structure is constructed, the carrier transport path is controlled based on the quantum confinement effect of the quantum well-quantum dot composite structure, the time-series data sequence is mapped into quantum states, and the energy level distribution of the quantum well-quantum dot composite structure is adjusted through quantum state engineering to form a quantum feature sequence. A quantum transmission channel is constructed using the quantum tunneling effect. The quantum transmission channel includes a quantum tunneling barrier array. The carrier transport is controlled by the Coulomb blocking effect of the quantum tunneling barrier array. A device data sequence is generated based on the coupling effect of the quantum tunneling effect and the quantum confinement effect. A long short-term memory network is established to parse the device data sequence. The output features of the long short-term memory network are correlated and mapped with the coherence characteristics of the quantum state. Based on the coherence function of the quantum state, the historical operating characteristics of the device are quantized to generate device feature identifiers.

3. The method according to claim 2, characterized in that, A quantum transmission channel is constructed using the quantum tunneling effect. The quantum transmission channel includes a quantum tunneling barrier array. Carrier transport is modulated through the Coulomb blockade effect of the quantum tunneling barrier array. A device data sequence is generated based on the coupling effect of the quantum tunneling effect and the quantum confinement effect, including: A periodically arranged quantum tunneling barrier array is constructed, and a gradient micro / nano structure is used to establish a carrier wave function in the barrier region of the quantum tunneling barrier array. The exponential wave vector parameter of the carrier wave function is modulated based on the gradient micro / nano structure. A Coulomb blocking effect control mechanism is constructed under the action of the current carrier wavefunction. The transmission coefficient is generated by enhancing the coupling effect of the left and right potential barriers using surface plasmon resonance. The tunneling current corresponding to the current carrier wavefunction is controlled based on the transmission coefficient. The Coulomb blocking effect control mechanism and the quantum confinement effect are coupled into a system. The coupled state transmission and uncoupled state transmission of the coupled system are realized by external field control, generating a coupling enhancement factor. The tunneling current is modulated according to the coupling enhancement factor to generate a device data sequence.

4. The method according to claim 1, characterized in that, Based on the Long Short-Term Memory network, the control command sequence of the core node layer is decoupled, and a mapping relationship between the device feature identifier and the control command is established to generate an initial control scheme, including: A long short-term memory network is trained to establish a temporal correlation model of device features. Based on the temporal correlation model, the control command sequence of the core node layer is decoupled. A gradient feedback adjustment mechanism is used to construct a mapping relationship between the device feature identifier and the control command. The gradient feedback adjustment mechanism adjusts the mapping weights in real time. An initial control scheme is generated based on the iterative optimization results of the mapping weights.

5. The method according to claim 1, characterized in that, The control scheme components are mapped using a family of hash functions, and the mapping results are input into a neural symbolic reasoning system. Based on the neural symbolic reasoning system and an expert knowledge base, an optimized verification sequence is generated, including: The control scheme is decomposed into multiple control components. The multiple control components are mapped using a family of hash functions to obtain initial mapping features. A component feature matrix is ​​constructed based on the initial mapping features. The degree of correlation between components is determined by similarity calculation. The component feature matrix is ​​adjusted according to the degree of correlation to generate enhanced mapping features. The enhanced mapping features are input into the neural symbolic reasoning system. The neural symbolic reasoning system extracts information from the enhanced mapping features to construct a feature priority sequence. The enhanced mapping features are reconstructed based on the feature priority sequence to obtain reconstructed features. The temporal variation pattern of the reconstructed features is analyzed to generate temporal correlation features. A knowledge fusion module is constructed, and a mapping relationship between the temporal correlation features and the expert knowledge base is established based on the knowledge fusion module. A feature expression space is constructed according to the mapping relationship, and the temporal correlation features are transformed into standard feature representations in the feature expression space. An optimized verification sequence is generated by adjusting the standard feature representations based on a balance factor.

6. The method according to claim 1, characterized in that, Based on the control sequence formed by the final control scheme, the intelligent control status of the solar acoustic and optical equipment group is analyzed in real time to construct cross-domain collaborative data; The generation of secure data by establishing a hash mapping for the cross-domain collaborative data based on the Bloom filter includes: The control sequence is chaotically mapped to obtain the device state iteration value. A chaotic mapping model is constructed based on the device state iteration value. The dynamic change law of the device state parameters is analyzed based on the chaotic mapping model. A parameter mapping sequence is generated based on the dynamic change law. The state change trend of the device group is analyzed according to the parameter mapping sequence. The state change trend reflects the collaborative evolution process of the device group under the action of the parameter mapping sequence. The operational stability of the device group is judged based on the collaborative evolution process. The control strategy is dynamically adjusted according to the operational stability using the chaotic mapping model. The control strategy is constructed using the chaotic mapping model, a hash verification space is established in the hash verification space, and the security verification information of the control strategy is transmitted along the verification path to generate security data.

7. A collaborative control and optimization system for a group of solar-powered acoustic and optical devices based on the Internet of Things, used to implement the method of any one of claims 1-6, characterized in that, include: The first unit is used to map the positional source data of the solar-powered acoustic and optical equipment group into multi-dimensional temporal features and construct the equipment data sequence; Long Short-Term Memory (LSTM) networks are used to analyze the feature correlation in the device data sequence, quantify the historical operating characteristics of the device, and label the device feature identifiers. The second unit is used to calculate the device trust value based on the temporal change trend of the device feature identifier, and to divide the solar-powered acoustic and optical device group into a core node layer and a cooperative node layer according to the distribution characteristics of the device trust value. The third unit is used to decouple the control command sequence of the core node layer based on the long short-term memory network, establish the mapping relationship between the device feature identifier and the control command, and generate an initial control scheme; establish a verification sequence of the initial control scheme according to the Bloom filter of the cooperative node layer, match the device feature identifier to update the device trust value, and form a final control scheme. The fourth unit is used to analyze the intelligent control status of the solar-powered acoustic and optical equipment group in real time based on the control sequence formed by the final control scheme, and construct cross-domain collaborative data; to establish a hash mapping of the cross-domain collaborative data according to the Bloom filter to generate security data, to use the Long Short-Term Memory Network to analyze the temporal characteristics of the security data and establish an association model with the device feature identifier, to form a cross-domain resource allocation scheme, and the cross-domain resource allocation scheme is converted into device execution instructions through control sequence mapping.

8. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 6.

Images