In order to make rational and efficient use of biomass resources, considering the operating cost and environmental cost of each microsource unit, an economic dispatching model of combined cooling heating and power microgrid based on improved biomass gasification was designed. In order to solve the problem that the sparrow search algorithm is easy to fall into the local optimum, an improved sparrow search algorithm(ISSA) was proposed to solve the proposed model. First, a sine chaos map was used to generate spatially evenly distributed early sparrow populations. Secondly, a mutually beneficial learning mechanism was added and a mutation strategy was introduced to enhance the information sharing and global search ability among individuals in this field. Finally, by comparing the iterative results of ISSA, SSA, gray wolf algorithm, whale algorithm and marine predator algorithm, it is proved that ISSA has good optimization effect and stability. Through the analysis of typical simulation cases, the effectiveness of the ISSA algorithm in solving the economic dispatching problem of combined cooling, heating and power microgrid is verified.

In order to explore the connection between brain and vision and improve the clarity and accuracy of brain activity reconstruction video, a new method called high quality electroencephalogram video reconstruction (HQEEGVR) was proposed to reconstruct video from EEG (electroencephalogram) signals. Firstly, the masking spatio-temporal frequency fusion network (MSTFFNet), a three-branch EEG feature extraction network, was proposed to extract brain activity information from EEG signals and dig deeper into the semantics behind brain activity changes, spatio-temporal frequency information was extracted at the same time. Secondly, cross-modal contrast learning was introduced to align EEG, text and image features for use in the generation stage. Then, a cascade video diffusion model was proposed, specifically, the stable diffusion model was used to generate reference video frames based on EEG features, and then the video frames were used as references, motion vectors were integrated, and the video diffusion model was introduced to capture the video time features. High quality videos were ultimately generated. The results show that the model performs well in the reconstruction of the subject, motion, color and semantics of the video. It can be seen that the EEG signal can be used to capture the visual and semantic information of the brain activity, so as to reconstruct the video with high fidelity and visual authenticity.

To address the issues of low path planning efficiency, poor obstacle avoidance capability, and low path quality of the RRT-Connect algorithm in complex environments, an improved RRT-Connect algorithm was proposed. Firstly, a bidirectional goal bias strategy was introduced to enhance the goal-directedness and path planning efficiency of the algorithm. Secondly, an obstacle avoidance optimization strategy was proposed to increase the algorithm's active obstacle avoidance capability and passage ability in complex environments. Finally, a path recombination strategy and a smoothing strategy were added to optimize the generated initial path, reducing path length and the number of turns, and improving path quality. The improved algorithm was compared with other algorithms in three complex environments using MATLAB. Simulation results show that the improved algorithm has less planning time, shorter path length, fewer sampling times, and a higher success rate of path planning, demonstrating the effectiveness of the improved algorithm in complex environments.

Aiming at the nonlinearity and multi-disturbance problems of the electric regulating valve control system in the actual production process, a control method based on the improved ant colony algorithm to optimize the single neuron PID (proportional integral derivative) was proposed and applied to the valve opening control. The self-learning and self-adaptive ability of the single-neuron network was used to achieve the online tuning of PID control parameters. The improved ant colony optimization algorithm was adopted to optimize the learning rate and neuron ratio coefficients in the single-neuron PID, which effectively overcomed the shortcomings of the single-neuron PID where the learning rate and neuron ratio coefficients could not achieve the expected control effect due to the empirical setting. The simulation comparison results show that, compared with the traditional PID, single neuron PID, and single neuron PID based on ant colony optimization algorithm optimization of the three control methods, the control method proposed overshoots the amount of reduction of 10.2%, 6.1%, and 1.8%, respectively. At the same time, the regulation time is correspondingly shortened by 0.22s, 0.07s, and 0.03s. It shows a stronger adaptive and anti-interference ability, which can make the valve opening control more stable and reliable.

Network intrusion detection systems (NIDS) are critical for maintaining cybersecurity. However, due to the complexity of network traffic data and the issue of class imbalance, existing detection models often exhibit high false alarm rates and insufficient detection accuracy for different types of attacks. To address these challenges, an imbalanced learning method for network intrusion detection, based on topological data analysis (TDA) and named TopoSMOTE, was proposed. This method aims to balance the training dataset by generating new minority class samples. The core of TopoSMOTE lied in constructing topological graphs to synthesize new samples. Firstly, the method used TDA to map the spatial relationships and connection patterns in network traffic data, forming a topological graph. Then, based on the topological graph, a minority class sample selection strategy was designed, which synthesized new data by selecting the nearest neighbor samples with topological relationships in a low-dimensional mapped space. Experiments were conducted on two imbalanced datasets. The experimental results show that the TopoSMOTE method achieves higher detection accuracy and lower false alarm rates compared to advanced oversampling methods and intrusion detection models.

In order to solve the problems of inaccurate dense target recognition and difficult detection of small targets in bird recognition, a bird recognition algorithm based on improved YOLOv8 was proposed. Firstly, in order to solve the problem of difficult dense object recognition, the multi-scale linear attention mechanism EfficientViT was used to replace the backbone network to realize the global receptive field and multi-scale learning, improve the performance and efficiency of the model, and improve the dense object recognition effect. Then, in order to solve the problem that it is difficult to detect small target birds and is prone to missed detection, an efficient multi-scale attention EMA (efficient multi-scale attention) mechanism was introduced to realize cross-dimensional aggregation features through channel recombination, so as to better capture global information, realize multi-scale feature fusion, and reduce the probability of missed detection. The experimental results show that the mAP50 of the improved model on the benchmark dataset CUB-200-2011 and birds28 reaches 77.1% and 88.4%, respectively, which is 4.5 and 5.4 percentage points higher than the original YOLOv8 model, respectively, which verifies the effectiveness of the improved model.

Aiming at the limitations of current intelligent traceability and authenticity identification systems in extracting multiple surface texture features (such as continuous, non-continuous, etc.) of automotive components, a micro-visual and neural network-based automotive parts anti-counterfeiting feature extraction and automatic matching algorithm was proposed. This algorithm integrated artificial intelligence-based automatic matching technology with micro-visual image processing and a neural network hybrid algorithm for anti-counterfeiting feature extraction and identification of automotive parts. Initially, the micro-visual feature images of the automotive component surfaces were processed with frequency-domain transformation, filtering, and noise reduction. Subsequently, the texture types (including continuous, non-continuous, and contour types) were determined based on the two-dimensional frequency-domain features. For each texture type, an appropriate algorithm was selected from the algorithm library to extract and analyze key attribute feature points. Finally, a deep learning framework was constructed, and a micro-visual feature recognition model for automotive parts was built, which was then matched with a priori feature libraries to complete classification and authenticity determination. Experimental results demonstrate that the proposed algorithm effectively extracts and identifies anti-counterfeiting features on the surface of automotive components, achieving a significant improvement in accuracy compared to traditional methods. Through matching with the a priori feature library, the algorithm accurately distinguishes between genuine and counterfeit components, providing reliable anti-counterfeiting verification results. This method effectively addresses the complexity of extracting various surface texture features of automotive parts, enhancing the accuracy of anti-counterfeiting and traceability systems. The micro-visual and neural network-based automatic matching technology significantly improves the precision of authenticity identification, offering an innovative and efficient solution for automotive parts anti-counterfeiting.

In order to explore the strengthening effect of reinforced concrete beams strengthened by carbon fiber reinforced polymer reinforced geopolymer matrix (FRGM), the mechanical properties of reinforced concrete beams reinforced with flexural reinforcement reinforced concrete in FRGM system were studied by numerical simulation. Three-dimensional finite element models were established to simulate the failure mode, characteristic load and load-mid-span deflection curve of reinforced concrete beams reinforced by FRGM system, and the influence of the thickness and length of the FRGM layer and the pre-damage degree of the original specimen on the reinforcement effect was discussed. The results show that the thickness of the FRGM layer has no obvious effect on the ultimate bearing capacity of reinforced concrete beam, and the increase range is 72.29%~79.38%, but the stiffness of the FRGM layer is improved to a certain extent, up to 38%. The ultimate bearing capacity of reinforced concrete beams can be increased by increasing the length of the FRGM layer, but with the increase of the length of the FRGM layer, the increase of the ultimate bearing capacity gradually weakens, indicating that there is no linear increase relationship between the length of the FRGM layer and the ultimate bearing capacity. Compared with the original components, the bearing capacity of reinforced concrete beam members is improved to a certain extent after pre-damage reinforcement(36.5%~73.66%), which shows the effectiveness of FRGM reinforcement method.

The slope geological structure characteristics is detected with the high density resistivity method by the inversion of the soil resistivity, and it could provide a geological model for slope stability analysis. However, the indirect evaluation of the resistivity for the soil shear strength is still limited. Taking the laterite on the slope as an example, the resistivity and shear strength of laterite samples with different dry density and water contents were tested to discuss the relationship between the resistivity and undrained shear strength of the laterite, and finally the corresponding quantitative model was established. The results show that the resistivity of laterite decreases with the increasing water content and increases with the increasing porosity. The undrained shear strength of laterite increases first and then decreases with the increasing water content (the peak of shear strength near the optimal water content ) and decreases with the increasing porosity. The evaluation model of undrained shear strength resistivity of unsaturated laterite considering critical saturation is derived, which is based on the three-phase conductivity theory of unsaturated soil and the shear strength theory of soil. The accuracy of the model is verified to be high, and there is a corresponding critical resistivity value for the peak change of undrained shear strength of laterite. As a physical parameter of soil, resistivity can be quickly detected and obtained. This model can provide new ideas for shear strength calibration, slope stability analysis and monitoring and early warning of laterite slope.

The stiffly-expanded composite pile is a new type of composite pile formed by sinking rigid pile in cement soil mixing pile. In order to study its bearing characteristics in complex soft soil-sandy soil foundation, a numerical analysis model of rigid composite pile was established based on finite element software ABAQUS, with the effect of core pile length and diameter, soil-cement pile length, diameter and elastic modulus, as well as soil-cement and pre-stressed high-strength concrete(PHC) pile interface and soil-soil interface shear strength on the bearing performance was analyzed. The results show that the bearing capacity and economic advantages of the stiffened composite pile are obvious. The bearing capacity of composite pile increases with the increase of research parameters, but the improvement is limited by changing the research parameters beyond a certain range. With the increase of load, the slope of axial force distribution curve of composite pile increases, and the load proportion of PHC core pile is about 93.32%~95.40%. When the load exceeds 4 000 kN, the axial force of soil-cement pile increases sharply within the depth range of 10~16 m, and the axial stress ratio of core pile/soil-cement pile changes significantly near the depth of about 10 m. The research outcomes could provide references for the engineering design and application of the stiffly-expanded composite piles.