Aiming at the influence of temperature cumulative effect on load change, a load forecasting algorithm for smart grid considering temperature cumulative effect was studied. The effect of continuous high temperature on grid load was included in the forecasting model. A modular neural network was used to ensure the independence of temperature cumulative effect learning and the improvement of accuracy. Three sub-networks formed the first layer of the multi-module neural network, with temperature, time and load characteristics as input parameters. The accuracy of load forecasting is 98.13％, and the error is 28.63％ lower than that before correction. The results show that the as-proposed algorithm has higher forecasting accuracy and operation efficiency.

To solve the problem of low transmission efficiency in wireless energy transmission of multi-degree-of-freedom motor, a wireless power transmission method based on magnetic coupling resonance was proposed. In accordance with the resonance principle, the wireless power transmission model based on magnetic coupling resonance was constructed, the impedance matching state was described by the reflection coefficient to obtain the optimal load impedance; the chaotic optimization algorithm was used to realize the optimization of coil loss rate and wireless power transmission for multi-degree-of-freedom motor. The experimental results show that the power output of the multi-degree-of-freedom motor reaches 893 W after using this method and the transmission efficiency is improved by more than 0.10. The power output and the transmission efficiency are enhanced simultaneously, indicating the feasibility and effectiveness of the as-proposed method.

During the identification of fragile lines in power grid by using pure cloud computation, there are problems such as high computing pressure on main nodes and high time consumption. By introducing the cloud edge coordination method, a method based on cloud edge coordination for identifying fragile lines in power grid was designed. Based on the cloud edge collaborative logic architecture, cloud and edge computations were integrated to build a fragile line identification model. The PageRank algorithm was used to establish DNS cache and integrate edge device data sources under cloud computation. Based on Kirchhoff′s law, the power flow values of different lines were calculated, and the identification of fragile lines was completed according to the power flow results. The results show that the as-proposed method has better performance and lower time consumption when identifying fragile lines in power grid.

Aiming at the problem that magnetic particle imaging (MPI) has been stayed in the laboratory mouse-scale stage, due to the conventional coils being incapable of constructing a high-strength gradient magnetic field for human use within an acceptable size, a method based on superconductor was proposed to develop miniaturized MPI system with a high-strength magnetic field. High-temperature superconductor MgB₂ coils were employed to replace the conventional copper coils to construct selection and drive fields of magnetic particles. The results show that under the condition of achieving the same order of magnetic flux, the total weight of two functional modules decreases to 0.01 kg, and the equipment volume decreases to 7 L due to the superconductor MPI. On this basis, the superconductor coils can provide a gradient magnetic field with a steeper slope of change, contributing to a clear and regular boundaries of field-free scanning region and more stable shape during the moving process with an average diameter narrows by about 50%.

Aiming at the problem of complex structure, diverse and uncertain data of the ubiquitous Internet of Things in electricity, an intelligent big data mining method based on multi-modal Markov decision for ubiquitous Internet of Things in electricity was proposed. This method constructed a Markov decision algorithm based on maximum entropy for the fault diagnosis and load forecast of ubiquitous Internet of Things, featuring small demand for labeled samples and high confidence. This method extracted power grid data characteristics by combining electrical quantity information and switch quantity information, and could make full use of multi-modal data samples. Simulation analysis and experimental results show that the as-proposed method can effectively identify grid faults including information distortion in comparison with traditional methods, and can effectively improve the accuracy of grid fault diagnosis and the precision of grid load forecasting.

Due to the increasing frequency of software release iterations, it is difficult to meet the needs of software delivery, thus the key technology for R＆D, operation and maintenance integration based on microservices were studied. The key technology for microservice publication was realized by coarse granularity, the virtualization technology was realized by container-based automatic discovery mechanism and microservice registration mechanism, and the integration of operation and maintenance was completed by using DevOps technology and agile development methodology. The experimental results show that the generation time of the as-proposed method is within 10 s, the reliability is maintained at more than 90％, and the memory occupied in the process of program execution fluctuates around 50 MB, which can effectively improve the operation efficiency of R＆D, operation and maintenance integration, shorten the deployment time, reduce the execution memory, ensure the reliability and standardization, and provide a preliminary reference for R＆D, operation and maintenance integration technology.

Aiming at the problem of low intelligence in the analysis and processing of infrared monitoring images for primary equipment within intelligent substations, a Faster R-CNN-based infrared detection image analysis method for abnormal equipment within substations was proposed to realize the intelligent identification and cause analysis of faulty equipment in substations. The infrared images collected by the remote terminal unit (RTU) were preprocessed, and the substation equipment in images was identified. The OSTU algorithm in combination with the image gray value was used to rectify and segment image. Faster R-CNN was used to distinguish whether the equipment was in fault state and analyze the cause. Experimental results show that the as-proposed method has a recognition accuracy of more than 90％ for a variety of faulty equipment, and has a comparatively excellent robustness.

To study the formation sequence and mechanism of intermetallic compounds in Al/Cu coatings, Al/Cu coating was prepared on 304 stainless steel substrate. Subsequently, the samples were heat treated to obtain intermetallic compounds by in-situ reaction within Al/Cu coating. In addition, the oxidation resistance of the coating at high temperatures was tested. First principles were used to calculate the enthalpy, entropy, Gibbs free energy and heat capacity of Al-Cu intermetallic compounds. Through the combination of thermodynamics with diffusion dynamics, an effective free energy of formation model was proposed to predict the formation sequence of compounds at the Al/Cu interface. The results show that the Al₂Cu phase is formed first within the Al/Cu coating, the AlCu phase appears at the Al/Cu interface, and the Cu-rich Al₄Cu₉ phase appears as the heating time increases to 20 h. The formation sequence of Al-Cu intermetallic compounds is Al₂Cu→AlCu→Al₄Cu₉, which is obtained by first principle calculation and is consistent with the experimental results. The oxidation experimental results show that the coating has excellent oxidation resistance at high temperatures.

In order to further improve the strength of high entropy alloy (HEA) with FCC structure, a Ni₃₈Co₂₅Cr₁₅Fe₁₀Al₁₀Ti₁Nb₁ HEA was prepared by vacuum arc melting method. The results show that the solidified structure of HEA is composed of single-phase FCC, and the typical dendrite with composition segregation can be observed within the as-cast strip. Besides, the HEA shows high recrystallization resistance, and the recrystallization fraction is only about 45.1％ at the annealing temperature of 1 000 ℃. In contrast, a completely recrystallized structure can be achieved when the annealing temperature increases to 1 050 ℃. Homogenous microstructure with refined grain size is obtained with the fully recrystallized HEA, and the corresponding average grain size is about 4.25 μm. Excellent tensile properties can be obtained with the annealed Ni₃₈Co₂₅Cr₁₅Fe₁₀Al₁₀Ti₁Nb₁ HEA. When annealing at 1 000 ℃, the ultimate tensile strength is about 1 640 MPa with an elongation of about 20％. For the HEA annealed at 1 050 ℃, the ultimate tensile strength and the elongation are about 1 580 MPa and 28％, respectively.

In order to study the influence of hole location and size on fatigue crack propagation path, ABAQUS finite element simulation software was used to establish the simulation model of crack propagation for modified CT samples. By using the extended finite element method (XFEM), the influence of the angle and distance between the circular hole and the crack tip as well as the radius of circular hole on the crack propagation path of Q235 steel under low cycle fatigue was studied. By introducing the crack deflection coefficient, the influence of different parameters on the deflection degree of crack propagation path was described quantitatively. The results show that the crack deflection coefficient reaches the maximum value when the angle between the crack tip and the circular hole is 45°. The crack deflection coefficient increases with the increase of hole radius, while decreases with the increase of the distance between circular hole and crack tip.

In view of the poor accuracy of label propagation in community network, a stable label propagation algorithm based on node influence value was proposed. The similarity of adjacent nodes in the community network of as-proposed algorithm was reset and transformed into an undirected and weighted graph of links from the perspective of link weighting. K-shell decomposition value of nodes was used to judge node influence, the centrality measure of seed nodes was calculated, and the update sequence of nodes was ranked in ascending order. According to the idea of the more important node with the label of higher influencing degree, the label of high frequency was modified to obtain the optimal result of community network division. The results show that the modular degree of as-proposed algorithm is above 0.320 and the accuracy reaches 99％. Therefore, the network structure divided by this method is clear, accurate and stable.

In order to solve the problem of low fitness value and long planning time resulting from the failure to obtain the target signal strength for intelligent robot in path planning, an intelligent robot path planning method based on bat optimization algorithm was proposed. The robot model was established to obtain its target signal strength, and the moving target was searched by particle swarm optimization. Bat algorithm and golden sine algorithm were combined to obtain the average position of population. The robot′s moving path was planned by searching process in stages. The results show that the path planning time of as-proposed method is only 2.0 s, the fitness is 24.1, and the number of infeasible solutions is zero. The as-proposed method can effectively improve the fitness value with reduced planning time, showing feasibility and value for practical application.

Aiming at the lower efficiency and accuracy of the existing Modbus TCP protocol for abnormal detection, a random forest abnormal detection model based on hybrid whale optimization algorithm was proposed. In this model, Cauchy mutation and adaptive dynamic inertia weight were combined. On the one hand, Cauchy mutation operator was used to increase population diversity and avoid the algorithm falling into local optimum. On the other hand, the adaptive dynamic inertia weight factor was used to improve the global search ability of the population, in order to increase the convergence speed of algorithm. Simulation results show that the as-proposed model has higher accuracy and stronger adaptability than other classification algorithms, indicating that the model also has higher detection accuracy in practical application.

Aiming at the problem of control accuracy of movement trajectory for intelligent storage unmanned vehicle, a path tracking control algorithm based on reinforcement learning of laser navigation unmanned vehicle was proposed. The mathematical model of intelligent storage unmanned vehicle was established, a laser navigation positioning algorithm was designed, and an Actor-Critic reinforcement learning path tracking algorithm was proposed. The high precision control of unmanned vehicle path tracking was realized. The comparative simulation verifies that the as-designed path tracking algorithm has better control effect and higher control accuracy. The maximum path tracking error is only 0.3 m; the maximum x-coordinate tracking error is only 0.1 m; the maximum y-coordinate tracking error is only 0.29 m, greatly improving the tracking accuracy of the motion trajectory of storage unmanned vehicle.

Aiming at the problem of low recognition rate and slow detection speed of magnetic flux leakage (MFL) defects, a MFL defect recognition method based on attention feature fusion was proposed. The algorithm was modified on the basis of CenterNet. A lightweight network PP-LCNet was selected as the backbone network, which could simultaneously guarantee low computation and high accuracy, compared with the popular backbone feature extraction network. At the same time, the attention network CBAM was used to positively learn the important information about the low-level features and integrate it with the high-level features. The model could obtain both low-level fine-grained information and high-level semantic information to improve the accuracy of small defect recognition. The experimental results show that the accuracy of as-proposed algorithm is 94.3％ and the inference time is 9.6 ms, respectively, when the IOU is greater than 0.5.

In order to recycle and utilize iron tailings, raw materials including iron tailings, fly ash and slag were used to prepare foam concrete for partial substitution of cement by alkali-excited cementitious materials. The effects of alkali-excited cementitious materials (cement substitution rate), foam, water-binder ratio and polyvinyl alcohol on the compressive strength and thermal conductivity of foam concrete were analyzed, and the microstructure and reaction mechanism were investigated by SEM and XRD. The test results show that the maximum 28 d compressive strength is 4.99 MPa, the dry apparent density is 784 kg/m³, and the thermal conductivity is 0.190 9 W/(m·K), when the content of alkali-excited cementitious material is 40％, the water-binder ratio is 0.40 and the polyvinyl alcohol is 1％, meeting the specification requirements.

Aiming at the difficulties of stockpiling, low comprehensive utilization rate and low activity, a kind of multi-solid waste concrete with iron tailings, steel slag and desulfurized ash as composite admixture was prepared. The effects of the content of composite admixtures and the fineness of iron tailings on the compressive strength of concrete were studied by testing the compressive properties, and the microstructure of concrete was investigated by mercury injection method (MIP) and backscattering electron imaging technology (BSE). The results show that the composite admixture significantly influences the early compressive strength of concrete. The 28 d compressive strength of concrete with less than 20％ admixture is the same as that of the test group without admixture. The compressive strength of concrete with 30％ admixture increases first and then decreases with the increase of the specific surface area of iron tailings. Adding composite admixtures and reducing the fineness of iron tailings can improve the pore structure of concrete and the compactness of the interfacial transition zone.

In order to solve the problems of higher construction cost and difficulty of fan foundation, a new anchor bolt inverted rib fan foundation was proposed. The finite element analysis software ABAQUS was used to analyze the concrete stress, reinforcement stress, foundation displacement and foundation inclination rate for the novel fan foundation and the traditional beam-plate fan foundation under extreme operating load. Furthermore, the concrete dosage, steel dosage and earthwork excavation quantity were discussed. The results show that the mechanical performance and economic efficiency of anchor bolt inverted rib fan foundation are obviously better than those of traditional beam-plate fan foundation, owing to the addition of anchor bolt and the design of inverted rib structure. The anchor bolt inverted rib fan foundation has a good application prospect.