4results about How to "Realize hierarchical control" patented technology

The application discloses a fan control circuit and an air conditioner. The fan control circuit is connected with a fan. The fan control circuit comprises a control module, a sampling module and a detection module. The control module outputs a fan current to control the start of the fan when the fan starts. The sampling module samples and amplifies the fan current to output an amplified signal. The detection module is used for outputting a first detection signal when the voltage of the amplified signal is less than or equal to a first threshold value, outputting a second detection signal when the voltage of the amplified signal is greater than the first threshold value and less than a second threshold value, and outputting a third detection signal when the voltage of the amplified signal is greater than or equal to the second threshold value. The control module further adjusts the fan current according to the first detection signal, the second detection signal or the third detection signal, can meet the start of the fan in different working modes, improves the test accuracy of the fan, and reduces the damage rate of the fan.

The embodiment of the invention provides a training method and platform for a diffusion language model, equipment, a medium and a product. According to the scheme, the method comprises the steps of obtaining a plurality of intermediate prediction results generated by an initial diffusion language model for sample cue words; for each of the plurality of intermediate prediction results, determining a reward value of the prediction result; and adjusting parameters of the initial diffusion language model based on the reward value to obtain a trained diffusion language model. According to the scheme, the disadvantage that the contribution degrees corresponding to the rewards and the intermediate prediction results are not matched can be effectively avoided, refined hierarchical regulation and control of the model optimization process can be realized, the overall training effect of the model can be improved, and the reasoning performance of the diffusion language model after training can be improved.

The present invention belongs to the field of heavy oil exploitation, and discloses a multi-layer injection and graded controllable injection device and method for injecting hot fluid. The device comprises a steam injection column and a graded injection device, the latter comprising a shell, the bottom wall of which is sealedly connected to the bottom wall of the steam injection column, the top wall of which is sealedly connected to the outer peripheral wall of the steam injection column, a distance is formed between the outer peripheral wall of the shell and the outer peripheral wall of the steam injection column to form a gas / steam injection chamber, which is axially divided into a plurality of independent sub-gas / steam injection chambers, adjacent sub-gas / steam injection chambers can be selectively connected, each sub-gas / steam injection chamber corresponds to a different oil layer, each sub-gas / steam injection chamber can be selectively connected to the corresponding oil layer, so as to inject gas / steam into a certain oil layer alone, or inject gas / steam into multiple oil layers at the same time, or inject gas / steam into oil layers distributed from bottom to top from bottom to top or from top to bottom in sequence, and a gas hole is formed in the area of ​​the sub-gas / steam injection chamber located at the bottom of the steam injection column.

The application provides a kind of based on multi-scale spectrum constraint's generated synthetic seismic data optimization method, comprising: actual seismic data and original synthetic seismic data are normalized and preprocessed;Actual seismic data is executed two-dimensional discrete wavelet transform, and low-frequency approximation sub-band and multi-stage high-frequency detail sub-band are obtained, and the condition input feature is constructed by channel splicing;Condition generative adversarial network is constructed, generator fuses multi-scale wavelet condition feature and generates optimization data, and discriminator carries out adversarial training;The same wavelet transform is executed to generated data, and multi-scale spectrum constraint loss is constructed, respectively matches low-frequency sub-band and each stage high-frequency detail sub-band, and establishes hierarchical spectrum constraint mechanism;Low frequency and high frequency constraint loss are weighted and fused, and are jointly optimized with adversarial loss to form total loss function, and generator parameter is iteratively updated;KL divergence is calculated to the optimized synthetic seismic data and actual data, and is filtered and output according to distribution similarity.