Intelligent energy-saving regulation and control method for circular flotation machine based on energy consumption model

Abstract

This invention discloses an intelligent energy-saving control method for circular flotation machines based on an energy consumption model, belonging to the field of intelligent energy-saving control technology for mineral flotation. The method includes acquiring multi-source energy consumption data such as the instantaneous value sequence of motor power, the sequence of froth layer height changes, and the sequence of pulp concentration during the operation of the circular flotation machine. A two-dimensional energy consumption field is constructed using an energy consumption feature field builder, with pulp concentration as the horizontal axis and motor power as the vertical axis. Dynamic energy consumption trajectory lines are plotted. State disturbance events are marked and associated based on the froth layer height changes. Using these disturbance events as driving points, a reconstruction engine completes adaptive mesh partitioning and elastic deformation, generating a reconstructed energy consumption field with nonlinear coordinate mapping. Based on this, intelligent energy-saving control commands are generated. This invention achieves field-based characterization and adaptive reconstruction of flotation machine energy consumption, improving control accuracy.

Description

Technical Field

[0001] This invention belongs to the field of intelligent energy-saving control technology for mineral flotation, specifically an intelligent energy-saving control method for circular flotation machines based on an energy consumption model. Background Technology

[0002] Existing energy consumption control technologies for circular flotation machines mostly adopt a single-parameter independent monitoring mode, collecting instantaneous values ​​of motor power, slurry concentration, and froth layer height data separately. Energy consumption status is judged through fixed threshold judgment and linear numerical analysis. Energy-saving control operations are executed by adjusting discrete parameters. Multi-source energy consumption data are only statistically analyzed and compared at single points. A unified representation form with multi-parameter coupling correlation has not been established. Energy consumption changes are only presented as independent parameter values, without a continuous and field-specific representation carrier.

[0003] Traditional energy consumption control schemes cannot form a two-dimensional correlation field between slurry concentration and motor power, nor can they plot dynamic energy consumption trajectories reflecting the coupled changes between the two. When the froth layer height exceeds the preset stable range, only a simple over-limit warning is given, without classifying the disturbance type based on the direction and amplitude of height change, and it is also impossible to accurately correspond the disturbance state to the energy consumption trajectory nodes. The energy consumption analysis uses a field model with a fixed coordinate structure, which lacks adaptive adjustment capability. Linear coordinate mapping cannot match the nonlinear energy consumption characteristics during flotation machine operation, and the fixed field structure cannot fit the energy consumption distribution changes in real-time operation, making it difficult to support the generation of precise control commands.

[0004] This invention requires constructing a two-dimensional energy consumption field with slurry concentration as the horizontal axis and motor power as the vertical axis, and drawing dynamic energy consumption trajectory lines. Based on the change in foam layer height, the trajectory point state disturbance event is marked and associated with its type. The state disturbance event is used as the driving point to perform adaptive meshing and elastic deformation on the two-dimensional energy consumption field, generating a reconstructed energy consumption field with nonlinear coordinate mapping relationship. Intelligent energy-saving control commands are generated based on the reconstructed energy consumption field. Summary of the Invention

[0005] This invention aims to solve at least one of the technical problems existing in the prior art;

[0006] Therefore, this invention proposes an intelligent energy-saving control method for circular flotation machines based on an energy consumption model, comprising:

[0007] Acquire multi-source energy consumption data during the operation of the circular flotation machine, including a sequence of instantaneous motor power values, a sequence of changes in froth layer height, and a sequence of slurry concentration.

[0008] The multi-source energy consumption data is input into a preset energy consumption feature field builder. The energy consumption feature field builder constructs a two-dimensional energy consumption field with slurry concentration as the horizontal axis and motor power as the vertical axis based on the instantaneous value sequence of motor power and the slurry concentration sequence. The dynamic energy consumption trajectory line reflecting the change of motor power under different slurry concentrations is drawn in the two-dimensional energy consumption field.

[0009] The dynamic energy consumption trajectory is characterized by the foam layer height change sequence. When the foam layer height exceeds the preset stable range, the corresponding dynamic energy consumption trajectory point in the two-dimensional energy consumption field is marked as a state disturbance event. The state disturbance event type is associated with the direction and magnitude of the foam layer height change.

[0010] Based on the spatial distribution characteristics of the dynamic energy consumption trajectory line and the state disturbance event markers, the energy consumption field reconstruction engine is driven. The energy consumption field reconstruction engine uses the state disturbance event as the deformation driving point to adaptively divide the spatial structure of the two-dimensional energy consumption field into grids and elastically deform it to generate a reconstructed energy consumption field with nonlinear coordinate mapping relationship.

[0011] Based on the reconstructed energy consumption field, intelligent energy-saving control commands for the circular flotation machine are generated.

[0012] Furthermore, the energy consumption feature field builder constructs a two-dimensional energy consumption field with slurry concentration as the horizontal axis and motor power as the vertical axis based on the instantaneous value sequence of motor power and the slurry concentration sequence, including:

[0013] The timestamps of the instantaneous power sequence of the synchronous motor and the slurry concentration sequence ensure a one-to-one correspondence between data points;

[0014] In a two-dimensional plane, the instantaneous value of slurry concentration is used as the horizontal axis value and the instantaneous value of motor power is used as the vertical axis value to determine the energy consumption state point at each sampling time.

[0015] Connecting the energy consumption state points at all sampling times in chronological order forms a dynamic energy consumption trajectory line that changes continuously in a two-dimensional plane. The coordinate space formed by the two-dimensional plane and the dynamic energy consumption trajectory line is the two-dimensional energy consumption field.

[0016] Further, the feature labeling of the dynamic energy consumption trajectory line based on the foam layer height change sequence includes:

[0017] Real-time comparison of each measured value of the foam layer height change sequence with the preset upper and lower limits of foam layer stability;

[0018] When the measured value of the foam layer height exceeds the upper limit of the foam layer stability, a positive foam disturbance event is marked at the energy consumption state point corresponding to the measured value of the foam layer height.

[0019] When the measured value of the foam layer height is lower than the lower limit of the foam layer stability, a negative foam disturbance event is marked at the energy consumption state point at the time corresponding to the measured value of the foam layer height.

[0020] Record the absolute value of the deviation between the measured value of the foam layer height and the steady-state boundary for each marked event, as the amplitude of the state disturbance event.

[0021] Furthermore, the step of driving the energy consumption field reconstruction engine based on the spatial distribution characteristics of the dynamic energy consumption trajectory line and the state disturbance event markers includes:

[0022] The distribution density of dynamic energy consumption trajectory lines in a two-dimensional energy consumption field is analyzed, and grid cells of different sizes are used for initial partitioning in dense and sparse regions of dynamic energy consumption trajectory lines.

[0023] Centered on each state disturbance event marker point, the deformation of the initial mesh element under local stretching or compression is calculated based on the type and magnitude of the state disturbance event.

[0024] Based on the calculated deformation, the coordinates of all grid nodes in the initial grid cell are iteratively adjusted until the dynamic energy consumption trajectory segments of the state disturbance event marked by the deformed grid are evenly distributed in the grid, thus completing the generation of the reconstructed energy consumption field.

[0025] Furthermore, based on the reconstructed energy consumption field, intelligent energy-saving control commands for the circular flotation machine are generated, including:

[0026] In the reconstructed energy consumption field, based on the distribution of the dynamic energy consumption trajectory line under the new nonlinear coordinates, the energy consumption cluster center and the energy consumption sparse channel are calculated. The energy consumption cluster center is defined by the region where the dynamic energy consumption trajectory line is densely concentrated, and the energy consumption sparse channel is defined by the path through which the dynamic energy consumption trajectory line is traversed at low density and without state disturbance event markers.

[0027] The coordinate pairs of slurry concentration and motor power corresponding to the energy consumption sparse channel are extracted as a set of candidate energy-saving operating points.

[0028] For each candidate energy-saving operating point in the candidate energy-saving operating point set, a pre-trained foam layer steady-state discrimination model is used for evaluation. The foam layer steady-state discrimination model takes the slurry concentration and motor power corresponding to the candidate energy-saving operating point as inputs and predicts the probability value that the foam layer height remains stable when the candidate energy-saving operating point is running.

[0029] Candidate energy-saving operating points with probability values ​​exceeding the steady-state threshold are selected, and a list of target energy-saving operating points is generated.

[0030] The list of target energy-saving operating points is input into the control command synthesizer, which maps each target energy-saving operating point to a motor speed control command and a slurry valve opening control command for the circular flotation machine.

[0031] Based on the motor speed control command and the slurry valve opening control command, the operating status of the circular flotation machine is adjusted to achieve intelligent energy-saving control of the circular flotation machine;

[0032] In the reconstructed energy consumption field, based on the distribution of the dynamic energy consumption trajectory line under the new nonlinear coordinates, the energy consumption clustering center and the energy consumption sparse channel are calculated, including:

[0033] Under the nonlinear coordinates of the reconstructed energy consumption field, the dynamic energy consumption trajectory line is redrawn;

[0034] The reconstructed energy consumption field is divided into multiple regular sub-regions, and the number of times the dynamic energy consumption trajectory line passes through each sub-region is counted to form a density distribution map;

[0035] The center point of a sub-region whose density value exceeds the high density threshold is identified as the energy consumption accumulation center;

[0036] Search outward from the energy consumption cluster center to find a continuous path that connects two energy consumption cluster centers and whose density values ​​of the sub-regions they pass through are all below the low density threshold. The continuous path is identified as an energy consumption sparse channel.

[0037] Furthermore, the step of extracting the corresponding slurry concentration and motor power coordinate pairs from the energy-sparse channels as a set of candidate energy-saving operating points includes:

[0038] Along each identified energy-sparse channel, multiple sampling locations are selected according to a preset equidistant sampling strategy;

[0039] For each sampling point, based on its nonlinear coordinates in the reconstructed energy consumption field, the slurry concentration value and motor power value corresponding to the sampling point in the original two-dimensional energy consumption field are solved by using the inverse coordinate mapping relationship of the energy consumption field reconstruction engine.

[0040] The coordinate pair consisting of the slurry concentration value and the motor power value is used as a candidate energy-saving operating point, and all candidate energy-saving operating points constitute the set of candidate energy-saving operating points.

[0041] Furthermore, the evaluation of each candidate energy-saving operating point in the candidate energy-saving operating point set using a pre-trained foam layer steady-state discrimination model includes:

[0042] The foam layer steady-state discrimination model is a machine learning model trained based on historical operating data. Its input features include slurry concentration, motor power, and interaction features between slurry concentration and motor power.

[0043] The slurry concentration value and motor power value of each candidate energy-saving operating point in the candidate energy-saving operating point set, as well as the interaction feature term calculated from the slurry concentration value and motor power value, are input into the foam layer steady-state discrimination model.

[0044] The foam layer steady-state discrimination model outputs a foam layer steady-state probability value between zero and one. The foam layer steady-state probability value represents the predicted probability that the height of the foam layer in the circular flotation machine will remain within a preset stable range under the given combination of slurry concentration and motor power.

[0045] Furthermore, the process of filtering out candidate energy-saving operating points whose probability values ​​exceed the steady-state threshold and generating a list of target energy-saving operating points includes:

[0046] Set a steady-state probability threshold for the foam layer, and compare the probability value output by the steady-state discrimination model of the foam layer with the steady-state probability threshold of the foam layer;

[0047] Retain all candidate energy-saving operating points whose probability values ​​are greater than or equal to the steady-state threshold;

[0048] The remaining candidate energy-saving operating points are sorted in ascending order according to their corresponding motor power values ​​to generate an ordered list of target energy-saving operating points.

[0049] Furthermore, the control command synthesizer maps each target energy-saving operating point to a motor speed control command and a slurry valve opening control command for the circular flotation machine, including:

[0050] Establish a mapping table in advance from motor power to motor speed, and a mapping table from slurry concentration to slurry valve opening;

[0051] For each operating point in the target energy-saving operating point list, the mapping relationship table from motor power to motor speed is queried based on its motor power value to obtain the corresponding target motor speed value;

[0052] For the same working point, the mapping relationship table between slurry concentration and slurry valve opening is consulted based on its slurry concentration value to obtain the corresponding target slurry valve opening value;

[0053] The target motor speed value and the target slurry valve opening value are combined to form a set of motor speed control commands and slurry valve opening control commands.

[0054] Furthermore, based on the motor speed control command and the slurry valve opening control command, the operating status of the circular flotation machine is adjusted, including:

[0055] The motor speed control command is sent to the motor driver of the circular flotation machine, and the slurry valve opening control command is sent to the electric regulating valve of the slurry pipeline;

[0056] The motor driver adjusts the motor to the target speed based on the received motor speed control command;

[0057] The electric regulating valve adjusts the valve to the target opening degree according to the received slurry valve opening control command, thereby switching the circular flotation machine to the target energy-saving operating point.

[0058] Compared with the prior art, the beneficial effects of the present invention are:

[0059] A two-dimensional energy consumption field is constructed with slurry concentration as the horizontal axis and motor power as the vertical axis. Dynamic energy consumption trajectories reflecting changes in motor power under different slurry concentrations are plotted in the two-dimensional energy consumption field. Based on the sequence of changes in foam layer height, state disturbance events are marked at the corresponding nodes of the energy consumption trajectory. The disturbance event types are associated with the direction and amplitude of changes in foam layer height, forming a coupled characterization carrier for multi-source energy consumption parameters. This fully presents the linkage relationship between slurry concentration and motor power. The dynamic energy consumption trajectory can continuously record the real-time evolution of energy consumption with slurry concentration. The state disturbance event markings can accurately correspond to the associated nodes of foam layer anomalies and energy consumption changes. The binding of disturbance type and foam layer change characteristics can clarify the triggering direction and change amplitude of energy consumption anomalies, realizing the synchronous visualization of energy consumption status and disturbance characteristics.

[0060] The energy consumption field reconstruction engine is driven by state disturbance events as deformation driving points. Adaptive meshing and elastic deformation operations are performed on the two-dimensional energy consumption field to generate a reconstructed energy consumption field with nonlinear coordinate mapping. Based on the reconstructed energy consumption field, intelligent energy-saving control commands for the circular flotation machine are generated. The adaptive meshing can adapt to the distribution differences of the field space under different disturbance intensities. The elastic deformation can correct the adaptation deviation of the original fixed coordinate field. The nonlinear coordinate mapping can fit the nonlinear change law of energy consumption during the operation of the circular flotation machine. The reconstructed energy consumption field can directly form the energy consumption space characteristics that match the real-time operating state. The control commands can be directly generated based on the reconstructed field space characteristics, maintaining a high degree of consistency with the actual energy consumption distribution and disturbance state of the flotation machine. Attached Figure Description

[0061] Figure 1 This is a flowchart illustrating the steps of the intelligent energy-saving control method for circular flotation machines based on an energy consumption model as described in this invention.

[0062] Figure 2 A diagram showing the original two-dimensional energy consumption field and dynamic energy consumption trajectory of a circular flotation machine;

[0063] Figure 3 A flowchart for an energy consumption field reconfiguration engine-driven process;

[0064] Figure 4 A graph showing the relationship between the height of the foam layer and disturbance events in a circular flotation machine;

[0065] Figure 5 A visualization of the energy consumption field for reconstructing a circular flotation machine. Detailed Implementation

[0066] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. 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.

[0067] See Figure 1 This invention provides an intelligent energy-saving control method for circular flotation machines based on an energy consumption model, the method comprising:

[0068] During the operation of the circular flotation machine, multi-source energy consumption data is collected in real time, including instantaneous motor power sequences, froth layer height variation sequences, and slurry concentration sequences. The acquired multi-source energy consumption data is input into a preset energy consumption feature field builder. This builder constructs a two-dimensional energy consumption field with slurry concentration as the horizontal axis and motor power as the vertical axis, based on the instantaneous motor power and slurry concentration sequences. Within this two-dimensional energy consumption field, a dynamic energy consumption trajectory line reflecting the continuous change in motor power under different slurry concentrations is plotted. The dynamic energy consumption trajectory line is characterized by the froth layer height variation sequence. When the froth layer height exceeds a preset stable range, a mark is made at the corresponding dynamic energy consumption trajectory point in the two-dimensional energy consumption field, recording it as a state disturbance event. Simultaneously, the specific type of event is associated with the direction of the froth layer height change and the magnitude of the exceedance. Based on the spatial distribution characteristics of the dynamic energy consumption trajectory line in the two-dimensional field and the marked state disturbance events, the energy consumption field reconstruction engine is driven. This engine uses the marked points of each state disturbance event as deformation driving points to adaptively mesh and elastically deform the original two-dimensional energy consumption field spatial structure, ultimately generating a reconstructed energy consumption field with a nonlinear coordinate mapping relationship. Based on the generated reconstructed energy consumption field, intelligent energy-saving control commands for the circular flotation machine are analyzed and calculated to guide the equipment to operate at a more energy-efficient and stable operating point.

[0069] In one embodiment of the present invention, the timestamps of the instantaneous motor power sequence and the slurry concentration sequence ensure a one-to-one correspondence between the instantaneous motor power value and the instantaneous slurry concentration value at each sampling moment. In a two-dimensional plane coordinate system, the instantaneous slurry concentration value is set as the abscissa and the instantaneous motor power value as the ordinate, thereby determining the energy consumption state point corresponding to each sampling moment in the two-dimensional plane. All energy consumption state points are connected sequentially according to time order to form a continuously changing dynamic energy consumption trajectory line in the two-dimensional plane. This dynamic energy consumption trajectory line and its corresponding two-dimensional plane coordinate space together constitute a two-dimensional energy consumption field characterizing the energy consumption state of the equipment operation.

[0070] In specific implementations, the timestamps of the synchronous motor power instantaneous value sequence and the slurry concentration sequence are implemented using a time alignment algorithm. This algorithm takes the original timestamps of both the motor power instantaneous value sequence and the slurry concentration sequence as input and outputs a motor power instantaneous value sequence and a slurry concentration sequence with unified timestamps, ensuring that each unified timestamp corresponds to one motor power instantaneous value and one slurry concentration instantaneous value. In some embodiments, when the sampling frequency of the motor power instantaneous value sequence differs from that of the slurry concentration sequence, a linear interpolation method is used to interpolate the timestamp of the low-frequency sampling sequence to the timestamp of the high-frequency sampling sequence, ensuring a one-to-one correspondence between the data points of the two sequences. In other embodiments, when the sampling frequencies of the motor power instantaneous value sequence and the slurry concentration sequence are the same but the timestamps have a fixed offset, a timestamp translation method is used for alignment to eliminate the offset and maintain the correspondence. The synchronized motor power instantaneous value sequence and the slurry concentration sequence contain the same number of data points, each representing a measurement value at the same sampling time, providing a consistent data foundation for subsequently constructing a two-dimensional energy consumption field.

[0071] A coordinate system is defined in a two-dimensional plane, where the horizontal axis represents the slurry concentration and the vertical axis represents the motor power. For each sampling moment, an energy consumption state point is determined by using the instantaneous value of the slurry concentration as the horizontal axis value and the instantaneous value of the motor power as the vertical axis value. The coordinates of the energy consumption state point are given by the formula:

[0072]

[0073] in: This represents the energy consumption state point at the i-th sampling time. This represents the instantaneous value of the slurry concentration at the i-th sampling time. This represents the instantaneous value of the motor power at the i-th sampling time. In specific implementation, the instantaneous value of the slurry concentration... Instantaneous motor power value derived from synchronized slurry concentration sequence The sequence of instantaneous motor power values ​​after synchronization.

[0074] After determining the energy consumption state points at all sampling times, these points are connected in chronological order. The connection is achieved by sequentially connecting energy consumption state points at adjacent sampling times with line segments, forming a continuously changing dynamic energy consumption trajectory in a two-dimensional plane. This dynamic energy consumption trajectory reflects the combined change path of slurry concentration and motor power over time. In practice, straight line segments are used for connection, but optionally, cubic spline interpolation can be used to generate a smooth trajectory to better reflect the changing trend. The coordinate space jointly formed by the two-dimensional plane coordinate system and the dynamic energy consumption trajectory constitutes the two-dimensional energy consumption field. This field provides a basic framework for visualizing the evolution of energy consumption states for subsequent analysis.

[0075] To illustrate the example scenario and provide data comparison, we consider the data acquisition of a circular flotation machine during a ten-minute operating cycle. The instantaneous motor power value sequence is sampled once per second, yielding 600 data points, while the slurry concentration sequence is sampled once every two seconds, yielding 300 data points. Before synchronization, the two sequences have different lengths and their timestamps do not perfectly correspond, leading to broken trajectory points when directly plotted as energy consumption status points. By linear interpolation, the slurry concentration sequence is interpolated to the once-per-second timestamp, resulting in 600 data point pairs after synchronization. Each data point pair includes one instantaneous slurry concentration value and one instantaneous motor power value. These 600 synchronized energy consumption status points are plotted in a two-dimensional plane and connected sequentially to form a continuous and smooth dynamic energy consumption trajectory line. Data comparison shows that the dynamic energy consumption trajectory line before synchronization exhibits gaps and breaks, while the synchronized dynamic energy consumption trajectory line is coherent and complete, accurately representing the energy consumption change process. In some embodiments, the time alignment algorithm also includes a data cleaning step to filter out outliers in the instantaneous motor power value sequence or slurry concentration sequence that exceed the sensor's measurement range, thereby improving data quality. Optionally, the synchronization process can be completed in real time within the local data acquisition module of the circular flotation machine, reducing the latency of data transmission to the host computer. It is understandable that accurate timestamp synchronization is crucial for constructing an accurate two-dimensional energy consumption field.

[0076] In one embodiment of the present invention, each measured value in the foam layer height change sequence is compared in real time with preset upper and lower limits of foam layer stability. When a measured foam layer height exceeds the upper limit of foam layer stability, a positive foam disturbance event is marked at the energy consumption state point corresponding to that measured value. When a measured foam layer height is below the lower limit of foam layer stability, a negative foam disturbance event is marked at the energy consumption state point corresponding to that measured value. The absolute value of the deviation between the measured foam layer height and the steady-state boundary value corresponding to each marked event is recorded, and this absolute value is used as the amplitude parameter of the state disturbance event.

[0077] In specific implementations, each measured value of the foam layer height change sequence is compared in real time with preset upper and lower limits of foam layer stability. These preset upper and lower limits are pre-set according to the steady-state operating range of the flotation process; for example, the upper limit is set to 450 mm and the lower limit to 350 mm. The foam layer height change sequence is acquired by an ultrasonic level gauge or laser rangefinder installed above the circular flotation tank at a fixed sampling period. Each foam layer height measurement is accompanied by a precise timestamp. The comparison operation is implemented through program logic, which cyclically reads the latest foam layer height measurement and compares it with the preset upper and lower limits to determine whether the foam layer height measurement is within the preset stable range. In some embodiments, the real-time comparison operation is performed in a programmable logic controller (PLC), which executes the comparison logic once per scan cycle. In other embodiments, the real-time comparison operation is implemented via a script in a host computer's data acquisition and monitoring control system. The script accesses the foam layer height measurement values ​​in a real-time database in a polling manner.

[0078] When the program logic detects that the measured foam layer height exceeds the upper limit of foam layer stability, a positive foam disturbance event is marked at the energy consumption state point corresponding to the measured foam layer height. This marking operation involves adding an event type attribute to the energy consumption state point at that moment in the data structure storing the dynamic energy consumption trajectory line, with the event type attribute value set to "positive foam disturbance". In specific implementation, the data structure of the energy consumption state point includes an x-coordinate, y-coordinate, timestamp, and event marker field. When a positive foam disturbance occurs, the event marker field is updated. When the program logic detects that the measured foam layer height is below the lower limit of foam layer stability, a negative foam disturbance event is marked at the energy consumption state point corresponding to the measured foam layer height, and the event marker field is updated to "negative foam disturbance". Optionally, the marking operation can also be graphically rendered on the visualization interface of the two-dimensional energy consumption field, using a highlighted specific icon or color at the corresponding energy consumption state point location to provide an intuitive prompt.

[0079] Record the absolute value of the deviation between the measured foam layer height and the steady-state boundary for each marked event; this value is used as the amplitude of the state disturbance event. The amplitude of the state disturbance event is calculated using the formula:

[0080]

[0081] in: Indicates the magnitude of the state disturbance event. This indicates the measured value of the foam layer height. This represents the corresponding steady-state boundary value. When labeled as a positive bubble disturbance event, the steady-state boundary value is... The upper limit value for stability of the foam layer is used; when a negative foam disturbance event is identified, the steady-state boundary value is... The lower limit of foam layer stability is adopted. The calculated amplitude of state disturbance events, along with the corresponding event type and timestamp information, are recorded and associated with the corresponding energy consumption state point. It can be understood that the amplitude of state disturbance events quantifies the degree to which the foam layer height deviates from the steady-state boundary.

[0082] To illustrate with specific examples and data comparisons, consider a continuous sequence of foam layer height measurements. This sequence contains data from five sampling points, with timestamps from T1 to T5, corresponding to measurements of 340 mm, 360 mm, 470 mm, 430 mm, and 330 mm, respectively. The preset upper limit for foam layer stability is 450 mm, and the lower limit is 350 mm. The real-time comparison process is as follows: At timestamp T1, the measured value of 340 mm is below the lower limit of 350 mm, marked as a negative foam disturbance event with an amplitude of |340-350|=10 mm. At timestamp T2, the measured value of 360 mm is between the lower limit of 350 mm and the upper limit of 450 mm, and no event is marked. At timestamp T3, the measured value of 470 mm exceeds the upper limit of 450 mm, marked as a positive foam disturbance event with an amplitude of |470-450|=20 mm. At timestamp T4, the measured value of 430 mm is within the stable range, and no event is marked. At timestamp T5, the measured value of 330 mm is below the lower limit of foam layer stability, marked as a negative foam disturbance event with an amplitude of |330-350|=20 mm. Data comparison shows that at timestamps T2 and T4, there are no special markings on the dynamic energy consumption trajectory line, while at the energy consumption state points corresponding to timestamps T1, T3, and T5, the markings and corresponding amplitude values ​​of negative foam disturbance events, positive foam disturbance events, and negative foam disturbance events will be carried respectively.

[0083] See Figure 2 This is a two-dimensional energy consumption field and dynamic energy consumption trajectory diagram of a circular flotation machine, intuitively presenting the dynamic relationship between pulp concentration and motor power. The energy consumption trajectory is densely distributed within the range of 20–30% pulp concentration and 12–18kW motor power, reflecting the typical operating conditions of the flotation machine. The trajectory exhibits multi-segment fluctuation characteristics, indicating a dynamic coupling relationship between changes in pulp concentration and motor power adjustment. Early sampling points are concentrated in the 20–25% concentration and 12–18kW power range. Later sampling points shift towards the 27–31% concentration and 13–18.5kW power range, reflecting the adjustment trend of process parameters over time. This diagram is the basic output for constructing the energy consumption feature field, providing raw data coordinates for subsequent disturbance event labeling and field reconstruction. The dense / sparse regions of the trajectory are directly used for subsequent adaptive grid division and energy consumption cluster center identification, serving as the core basis for energy-saving path mining.

[0084] In one embodiment of the present invention, see [reference] Figure 3 This paper analyzes the spatial distribution density of dynamic energy consumption trajectories in a two-dimensional energy consumption field. Smaller grid cells are used for initial partitioning in areas with dense dynamic energy consumption trajectory distribution, while larger grid cells are used in areas with sparse distribution. Centered on each state disturbance event marker, the deformation of the event on the surrounding initial grid cells is calculated based on the event type and recorded amplitude parameters. Based on the calculated deformation, the coordinates of all grid nodes in the initial grid partitioning are iteratively adjusted until the distribution of dynamic energy consumption trajectory segments marked with state disturbance events becomes relatively uniform within the deformed grid. At this point, the generation of the reconstructed energy consumption field is complete.

[0085] In practical implementation, the distribution density of dynamic energy consumption trajectories in the two-dimensional energy consumption field is analyzed. The distribution density is calculated by counting the number of unit grids traversed by the dynamic energy consumption trajectory lines. A unit grid is a predefined, fixed-size square region covering the entire two-dimensional energy consumption field. The statistical process traverses all segments of the dynamic energy consumption trajectory lines, calculating the number of times each unit grid is traversed by a segment; this number is defined as the density value of that unit grid. Regions with high density values ​​are identified as dense dynamic energy consumption trajectory line regions, while regions with low density values ​​are identified as sparse dynamic energy consumption trajectory line regions. Initial mesh generation is performed based on the density analysis results. Smaller mesh cells are used for initial generation in dense dynamic energy consumption trajectory line regions, while larger mesh cells are used in sparse dynamic energy consumption trajectory line regions. In some embodiments, the initial meshing employs an adaptive quadtree algorithm, which recursively divides the two-dimensional energy consumption field into mesh cells. The termination condition for meshing is based on whether the sum of the lengths of the dynamic energy consumption trajectory line segments within the mesh cell exceeds a preset threshold. If the sum of lengths is large, the mesh cells are further subdivided; if the sum of lengths is small, subdivision ceases. This generates small-sized mesh cells in dense regions and large-sized mesh cells in sparse regions. In other embodiments, the initial meshing can also be based on the kernel density estimation results of the dynamic energy consumption trajectory line point set, dividing regions surrounded by high-density contour lines into small-sized meshes and regions surrounded by low-density contour lines into large-sized meshes.

[0086] Centered on each state disturbance event marker, the deformation of the initial mesh element due to local stretching or compression is calculated based on the type and magnitude of the disturbance event. In practice, the deformation calculation operates on the mesh element containing the state disturbance event marker and the nodes of its adjacent mesh elements. The calculation process considers the type of state disturbance event: positive foam disturbance events indicate that a local area needs to be stretched, and negative foam disturbance events indicate that a local area needs to be compressed. The magnitude of the deformation is directly proportional to the magnitude of the state disturbance event and inversely proportional to the mesh element size, ensuring that large mesh elements are insensitive to deformation changes, while small mesh elements can respond to fine changes. The deformation is calculated using the following formula:

[0087]

[0088] in: This represents the deformation caused by localized stretching or compression; a positive value indicates stretching, and a negative value indicates compression. Influence factors representing the types of state disturbance events, and influence factors corresponding to positive bubble disturbance events. A positive value indicates the impact factor corresponding to a negative bubble disturbance event. Negative value Indicates the magnitude of the state disturbance event. This represents the size of the initial mesh cell containing the state disturbance event marker. The calculated deformation... This will be applied to the coordinate adjustment of the grid cell containing the state disturbance event marker and its neighboring grid nodes. Optional, influence factor. The specific values ​​can be preset based on process knowledge, such as setting the influence factor of positive foam disturbance events. The impact factor of negative bubble disturbance events is 1.2. It is -0.8.

[0089] Based on the calculated deformation, the coordinates of all grid nodes in the initial grid cell are iteratively adjusted. The deformation calculation uses a physics-based spring-mass model to simulate the elastic deformation of the grid, where grid nodes are simulated as mass points, and the edges of the grid cells are simulated as springs. The deformation applied by the state disturbance event marker is converted into an external force on the surrounding mass points, driving the spring-mass system to a new equilibrium position. Each iteration recalculates the resultant force on each mass point and updates its position. The iteration terminates when the distribution of the dynamic energy consumption trajectory segments marked with state disturbance events in the deformed grid tends to be uniform. The uniformity of distribution is evaluated by calculating the length of the marker segments in each final grid cell and calculating the variance of these lengths. When the variance is lower than a preset threshold, the distribution is considered uniform, and the generation of the reconstructed energy consumption field is completed. It can be understood that the iterative adjustment stretches or compresses the grid near the state disturbance event, and the local shape of the dynamic energy consumption trajectory lines near the event points undergoes nonlinear changes in the reconstructed field, creating a new spatial structure for subsequent extraction of energy consumption features.

[0090] To combine specific example scenarios and data comparisons, it is assumed that a simplified two-dimensional energy consumption field initially contains two sizes of grid cells after grid division: the grid cell size in the dense region of the dynamic energy consumption trajectory is 1, and the grid cell size in the sparse region is 5. There are three state disturbance event markers in this field, with the following locations and event information: Event point 1 (coordinates (20,30), positive foam disturbance event, amplitude 20 mm), located in the small-sized grid area; Event point 2 (coordinates (150,80), negative foam disturbance event, amplitude 15 mm), located in the large-sized grid area. The influence factor F is set to +1.2 for positive events and -0.8 for negative events. According to formula (3), for event point 1, the deformation... A positive value indicates stretching; for event point 2, the deformation... Negative values ​​indicate compression. Data comparison shows that, at the same amplitude, event points located in small-sized grid cells produce much greater deformation than event points located in large-sized grid cells. This leads to more significant local grid deformation caused by state disturbance events in dense regions. In some embodiments, before iterative adjustment, the variance of the length distribution of the dynamic energy consumption trajectory segments marked with state disturbance events in the grid is 2.5. After multiple iterations of the spring-mass model, the grid node coordinates are adjusted, and the variance of the marked segments drops to 0.1, which is below the preset threshold of 0.15, at which point the iteration stops. Optionally, the spring stiffness coefficient in the spring-mass model can be set to be inversely proportional to the initial grid cell size to achieve adaptive deformation response. It can be understood that through this adaptive grid partitioning and elastic deformation, the reconstructed energy consumption field amplifies the characteristics of the state disturbance event region and makes the distribution of the dynamic energy consumption trajectory lines more uniform throughout the field, facilitating the identification of feature patterns by subsequent analysis algorithms.

[0091] In one embodiment of the present invention, the dynamic energy consumption trajectory line is redrawn under the nonlinear coordinates of the reconstructed energy consumption field. The entire reconstructed energy consumption field is divided into multiple regular sub-regions, and the number of times the dynamic energy consumption trajectory line passes through each sub-region is counted to form a density distribution map. The center point of the sub-region with a density value exceeding a preset high-density threshold is determined as the energy consumption cluster center. Starting from one energy consumption cluster center, a continuous path is searched outward to reach another energy consumption cluster center, and all the sub-regions traversed have a density value lower than a preset low-density threshold. This path is identified as an energy consumption sparse channel. Along each identified energy consumption sparse channel, multiple sampling point locations are selected according to a preset equidistant sampling strategy. For each sampling point location, based on its nonlinear coordinates in the reconstructed energy consumption field, the slurry concentration value and motor power value corresponding to the point in the original two-dimensional energy consumption field are solved using the inverse coordinate mapping relationship of the energy consumption field reconstruction engine. The coordinate pair formed by the solved slurry concentration value and motor power value is used as a candidate energy-saving operating point, and all such points constitute a set of candidate energy-saving operating points. The slurry concentration value and motor power value of each point in the candidate energy-saving operating point set, along with the interaction feature term calculated from them, are input into a pre-trained foam layer steady-state discrimination model. This model outputs a foam layer steady-state probability value between zero and one. A foam layer steady-state probability threshold is set, and the probability value output by the model is compared with this threshold. All candidate energy-saving operating points with probability values ​​greater than or equal to this steady-state threshold are retained. These retained operating points are then sorted in ascending order according to their corresponding motor power values ​​to generate an ordered list of target energy-saving operating points.

[0092] In practical implementation, the dynamic energy consumption trajectory is redrawn under the nonlinear coordinates of the reconstructed energy consumption field. The mapping relationship between the original coordinate point sequence of the stored dynamic energy consumption trajectory and the coordinates of the reconstructed energy consumption field is completed. This mapping relationship is established by the energy consumption field reconstruction engine during the deformation process. Each original coordinate point calculates its new coordinates in the reconstructed energy consumption field through the mapping relationship, and these new coordinate points are connected sequentially to form the redrawn dynamic energy consumption trajectory under the nonlinear coordinates. The reconstructed energy consumption field is divided into multiple regular sub-regions, based on the boundary range of the reconstructed energy consumption field. For example, the horizontal and vertical axes are divided into N and M equal segments respectively, forming N multiplied by M rectangular sub-regions. The number of times the dynamic energy consumption trajectory passes through each sub-region is counted. This counting process iterates through each segment of the redrawn dynamic energy consumption trajectory, determining which sub-regions the segment passes through and incrementing the count of the corresponding sub-region. After all segments have been traversed, a density distribution map is generated. In practical implementation, the density value of a sub-region is the ratio of the total length of the dynamic energy consumption trajectory line segments passing through that sub-region to the area of ​​the sub-region. Optionally, the density value can also be simplified to the number of times the dynamic energy consumption trajectory line segments pass through the sub-region. The center point of a sub-region whose density value exceeds a high-density threshold is identified as an energy consumption cluster center. The high-density threshold is set based on the statistical quantile of the density value distribution of all sub-regions; for example, sub-regions with the highest density values ​​(top 10%) are considered high-density regions. A search is conducted outward from one energy consumption cluster center to find a continuous path connecting two energy consumption cluster centers where the density values ​​of all traversed sub-regions are below a low-density threshold. The low-density threshold is set based on the statistical quantile of the density value distribution; for example, sub-regions with density values ​​below the 30th percentile are considered low-density regions. A continuous path satisfying the above conditions is identified as an energy consumption sparse channel. An energy consumption sparse channel connects two energy consumption cluster centers, and the path is entirely composed of low-density sub-regions.

[0093] Along each identified sparse energy consumption channel, multiple sampling points are selected according to a preset equidistant sampling strategy. "Equal spacing" means selecting points at fixed intervals along the centerline of the sparse energy consumption channel in the nonlinear coordinate space of the reconstructed energy consumption field. For each sampling point, based on its nonlinear coordinates in the reconstructed energy consumption field, the corresponding slurry concentration and motor power values ​​in the original two-dimensional energy consumption field are deduced through the inverse coordinate mapping relationship of the energy consumption field reconstruction engine. The inverse coordinate mapping relationship is the reverse process of coordinate transformation recorded by the energy consumption field reconstruction engine during deformation. The coordinate pair formed by the deduced slurry concentration and motor power values ​​is used as a candidate energy-saving operating point. All candidate energy-saving operating points constitute a set of candidate energy-saving operating points. For example, three sampling points are taken at equal intervals along an identified sparse energy consumption channel, and their corresponding original coordinates are obtained through inverse mapping, forming three candidate energy-saving operating points (see Table 1).

[0094] Table 1: Example Table of Candidate Energy-Saving Operating Point Sets

[0095] Sampling location points Reverse ore pulp concentration (%) Inversely solve for the motor power value (kW) Point 1 35.2 132.5 Point 2 36.8 130.1 Point 3 38.5 128.7

[0096] In some embodiments, the spacing of the equally spaced sampling is dynamically determined based on the length of the energy-sparse channel; longer channels have more sampling points, while shorter channels have fewer. Optionally, the sampling location can also be selected as the geometric center point of each low-density sub-region that makes up the energy-sparse channel.

[0097] The foam layer steady-state discrimination model is a machine learning model trained on historical operating data. Its input features include slurry concentration, motor power, and an interaction feature term between slurry concentration and motor power. The interaction feature term is typically the product of the slurry concentration value and the motor power value. Historical operating data contains a large number of foam layer height sequences under different combinations of slurry concentration and motor power, along with labels indicating whether the corresponding foam layer height is stable. The slurry concentration value and motor power value for each candidate energy-saving operating point in the candidate energy-saving operating point set, along with the interaction feature term calculated from these values, are input into the foam layer steady-state discrimination model. The model outputs a foam layer steady-state probability value between zero and one. This probability value represents the predicted probability that the foam layer height of the circular flotation machine will remain within a preset stable range under the given slurry concentration and motor power combination. The foam layer steady-state probability value is calculated using the following formula:

[0098]

[0099] in: This represents the steady-state probability value of the foam layer. This represents the slurry concentration value input to the steady-state discrimination model of the foam layer. This represents the motor power value input to the steady-state discrimination model of the foam layer. This represents the interaction characteristic term between slurry concentration and motor power. The prediction function of the foam layer steady-state discrimination model is trained by a machine learning algorithm. A steady-state probability threshold for the foam layer is set, pre-defined based on the process stability requirements, for example, 0.85. The probability values ​​output by the foam layer steady-state discrimination model are compared with the steady-state probability threshold, retaining all candidate energy-saving operating points with probability values ​​greater than or equal to the threshold. This comparison operation filters out operating points that, although located in energy-sparse channels, have poor predictive stability. The retained candidate energy-saving operating points are then sorted in ascending order according to their corresponding motor power values, generating an ordered list of target energy-saving operating points. The ordered list prioritizes operating points with lower motor power to achieve energy savings.

[0100] To illustrate with specific examples and data comparisons, we assume the reconstructed energy consumption field is divided into 9 sub-regions (3 rows, 3 columns). The density values ​​of each sub-region (simplified by the number of passes) are [15, 2, 18, 3, 1, 4, 20, 5, 16]. A high-density threshold of 10 is set, and sub-regions with density values ​​exceeding 10 are identified as center points (1,1), (1,3), (3,1), and (3,3), thus defining them as energy consumption cluster centers. A low-density threshold of 5 is set, and sub-regions with density values ​​below 5 are considered low-density regions. Searching for paths connecting the energy consumption cluster centers (1,1) and (1,3), we find that the path passes through sub-region (1,2), whose density value is 2, lower than 5. Therefore, the path [(1,1),(1,2),(1,3)] is identified as a sparse energy consumption channel. Data comparison shows that high-density regions reflect areas where equipment operates frequently, while low-density channels connecting high-density regions indicate potential transitional or low-energy-consumption state paths. Equally spaced samples were taken along this channel, and three candidate points were obtained through inverse mapping, as shown in Table 1. The data from these three points were input into the foam layer steady-state discrimination model. Assuming the calculated steady-state probability values ​​for the foam layer were 0.92, 0.87, and 0.81, respectively, and a steady-state threshold of 0.85 was set. After comparison, the point with a probability value of 0.81 was filtered out, and the two points with probability values ​​of 0.92 and 0.87 were retained. Arranged in ascending order of motor power value, the point with a motor power value of 130.1kW was placed before the point with a motor power value of 132.5kW. The generated list of target energy-saving operating points was in the order of (36.8%, 130.1kW), (35.2%, 132.5kW).

[0101] See Figure 4 This is a graph analyzing the relationship between the foam layer height and disturbance events in a circular flotation machine, used to analyze the correlation between foam layer height stability and energy consumption disturbances. Positive disturbance events are more numerous, concentrated around time points 0–30, 40, 55, 70, 85, and 95, indicating that excessive foam layer expansion is a more common type of instability. Negative disturbance events are mainly distributed at time points 5–15, 35–45, 50–75, and 90–100, corresponding to insufficient foam layer height. Most of the time, the foam layer height fluctuates between the upper and lower limits of stability, but there are several brief instances of exceeding the threshold, reflecting the dynamic characteristics of the flotation process. A significant deep negative disturbance occurs at time point 65–70, representing the most severe instability event observed. Disturbance events are mapped to the energy consumption field, serving as deformation driving points for energy consumption field reconstruction, used to optimize mesh generation and energy-saving path identification. The amplitude and frequency of positive / negative disturbances directly affect the steady-state probability assessment of subsequent candidate energy-saving operating points.

[0102] In one embodiment of the present invention, a mapping table from motor power to motor speed and a mapping table from slurry concentration to slurry valve opening are pre-established. For each operating point in the target energy-saving operating point list, the corresponding target motor speed value is obtained by querying the mapping table from motor power to motor speed based on its motor power value. For the same operating point, the corresponding target slurry valve opening value is obtained by querying the mapping table from slurry concentration to slurry valve opening based on its slurry concentration value. The obtained target motor speed value and target slurry valve opening value are combined to form a set of motor speed control commands and slurry valve opening control commands. The motor speed control commands are sent to the motor driver of the circular flotation machine, and the slurry valve opening control commands are sent to the electric regulating valve of the slurry pipeline. The motor driver adjusts the motor to the target speed according to the received motor speed control commands. The electric regulating valve adjusts the valve to the target opening according to the received slurry valve opening control commands, thereby switching the operating state of the circular flotation machine to the target energy-saving operating point.

[0103] In specific implementations, mapping tables from motor power to motor speed and from slurry concentration to slurry valve opening are pre-established. These two mapping tables are generated based on the equipment characteristics and historical calibration data of the circular flotation machine. The mapping tables are stored in the form of data pairs. The mapping table from motor power to motor speed contains multiple sets of motor power values ​​and corresponding target motor speed values, and the mapping table from slurry concentration to slurry valve opening contains multiple sets of slurry concentration values ​​and corresponding target slurry valve opening values. In some embodiments, the mapping relationships are described by linear interpolation functions, with function parameters obtained through equipment factory testing or on-site commissioning. In other embodiments, the mapping tables are stored in the non-volatile memory of the control system as lookup tables, supporting the query for the closest entry by input value or the calculation of output values ​​through interpolation. Optionally, the mapping relationships may also be non-linear, defined by piecewise linear functions or polynomial fitting curves, but ultimately, they are all available for real-time querying in the form of discretized data tables.

[0104] For each operating point in the target energy-saving operating point list, the mapping relationship table from motor power to motor speed is consulted based on the motor power value of the target energy-saving operating point to obtain the corresponding target motor speed value. The query process uses a binary search method to locate the key value in the mapping relationship table that is closest to the input motor power value and returns the corresponding target motor speed value. If the mapping relationship table supports interpolation, linear interpolation calculation is performed using two adjacent entries. For the same target energy-saving operating point, the mapping relationship table from slurry concentration to slurry valve opening is consulted based on the slurry concentration value of the target energy-saving operating point to obtain the corresponding target slurry valve opening value. The query process uses the same method as querying the motor speed. In specific implementation, the acquisition of the target motor speed value and the target slurry valve opening value is performed synchronously or sequentially. For a given operating point coordinate (slurry concentration value, motor power value), the output is the calculated pair of values ​​for the target motor speed and the target slurry valve opening. It can be understood that the mapping relationship table realizes the conversion from optimized operating point parameters to directly executable control parameters.

[0105] The target motor speed value and the target slurry valve opening value are combined to form a set of motor speed control instructions and slurry valve opening control instructions. The motor speed control instruction is an instruction in digital or analog signal format containing the target motor speed value, and the slurry valve opening control instruction is an instruction in digital or analog signal format containing the target slurry valve opening value. In some embodiments, the combination of control instructions is implemented by creating a structure or data object containing two fields: the target motor speed value and the target slurry valve opening value, with additional timestamp and operating point identification information. In other embodiments, the control instructions are combined in the form of specific communication protocol messages, for example, writing the target motor speed value and the target slurry valve opening value to different holding register addresses in a Modbus TCP message. Optionally, the mapping relationship can be expressed by the formula:

[0106]

[0107]

[0108] in: This indicates the target motor speed value obtained from the query. This represents the mapping function from motor power to motor speed, and its specific form is defined by the mapping table. This represents the motor power value at the target energy-saving operating point. This indicates the target slurry valve opening value obtained from the query. This represents the mapping function from slurry concentration to slurry valve opening; its specific form is defined by the mapping table. This indicates the slurry concentration value at the target energy-saving operating point.

[0109] The motor speed control command is sent to the motor driver of the circular flotation machine, and the slurry valve opening control command is sent to the electric regulating valve in the slurry pipeline. This sending operation is completed via an industrial fieldbus or analog output module. Based on the received motor speed control command, the motor driver adjusts the motor to the target speed. The motor driver uses an internal speed loop control algorithm to change the output frequency or voltage, ensuring the actual motor speed tracks the target motor speed. Based on the received slurry valve opening control command, the electric regulating valve adjusts the valve to the target opening. The positioning controller of the electric regulating valve drives the valve core to move until the valve opening feedback value equals the target slurry valve opening value, thus switching the circular flotation machine to the target energy-saving operating point.

[0110] See Figure 5 This is a visualization of the reconstructed energy consumption field of a circular flotation machine, primarily used to show the distribution characteristics of energy consumption trajectories under nonlinear coordinates. The more yellow the colored contour surface, the denser the energy consumption trajectory in that area; the more purple, the sparser. The white polyline is the motor power-slurry concentration change trajectory redrawn under nonlinear coordinates after adaptive grid deformation. The red pentagrams, defined by high-density trajectory convergence areas, represent core energy consumption hotspots and high-energy-consuming areas that should be avoided in energy-saving optimization. The density heatmap visually locates the energy consumption aggregation center, providing targeted optimization targets for energy-saving control. The reconstructed energy consumption trajectory clearly shows the changing patterns of equipment operating status in the nonlinear field, facilitating the discovery of potential energy-saving channels. It indirectly reflects the grid deformation effect driven by foam layer disturbance events, verifying the adaptive capability of the field reconstruction algorithm to disturbances.

[0111] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A method for intelligent energy-saving control of circular flotation machines based on an energy consumption model, characterized in that, include: Acquire multi-source energy consumption data during the operation of the circular flotation machine, including a sequence of instantaneous motor power values, a sequence of changes in froth layer height, and a sequence of slurry concentration. The multi-source energy consumption data is input into a preset energy consumption feature field builder. The energy consumption feature field builder constructs a two-dimensional energy consumption field with slurry concentration as the horizontal axis and motor power as the vertical axis based on the instantaneous value sequence of motor power and the slurry concentration sequence. The dynamic energy consumption trajectory line reflecting the change of motor power under different slurry concentrations is drawn in the two-dimensional energy consumption field. The dynamic energy consumption trajectory is characterized by the foam layer height change sequence. When the foam layer height exceeds the preset stable range, the corresponding dynamic energy consumption trajectory point in the two-dimensional energy consumption field is marked as a state disturbance event. The state disturbance event type is associated with the direction and magnitude of the foam layer height change. Based on the spatial distribution characteristics of the dynamic energy consumption trajectory line and the state disturbance event markers, the energy consumption field reconstruction engine is driven. The energy consumption field reconstruction engine uses the state disturbance event as the deformation driving point to adaptively divide the spatial structure of the two-dimensional energy consumption field into grids and elastically deform it to generate a reconstructed energy consumption field with nonlinear coordinate mapping relationship. Based on the reconstructed energy consumption field, intelligent energy-saving control commands for the circular flotation machine are generated.

2. The intelligent energy-saving control method for circular flotation machines based on an energy consumption model according to claim 1, characterized in that, The energy consumption feature field constructor constructs a two-dimensional energy consumption field with slurry concentration as the horizontal axis and motor power as the vertical axis based on the instantaneous value sequence of motor power and the slurry concentration sequence, including: The timestamps of the instantaneous power sequence of the synchronous motor and the slurry concentration sequence ensure a one-to-one correspondence between data points; In a two-dimensional plane, the instantaneous value of slurry concentration is used as the horizontal axis value and the instantaneous value of motor power is used as the vertical axis value to determine the energy consumption state point at each sampling time. Connecting the energy consumption state points at all sampling times in chronological order forms a dynamic energy consumption trajectory line that changes continuously in a two-dimensional plane. The coordinate space formed by the two-dimensional plane and the dynamic energy consumption trajectory line is the two-dimensional energy consumption field.

3. The intelligent energy-saving control method for circular flotation machines based on an energy consumption model according to claim 2, characterized in that, The feature labeling of the dynamic energy consumption trajectory line based on the foam layer height change sequence includes: Real-time comparison of each measured value of the foam layer height change sequence with the preset upper and lower limits of foam layer stability; When the measured value of the foam layer height exceeds the upper limit of the foam layer stability, a positive foam disturbance event is marked at the energy consumption state point corresponding to the measured value of the foam layer height. When the measured value of the foam layer height is lower than the lower limit of the foam layer stability, a negative foam disturbance event is marked at the energy consumption state point at the time corresponding to the measured value of the foam layer height. Record the absolute value of the deviation between the measured value of the foam layer height and the steady-state boundary for each marked event, as the amplitude of the state disturbance event.

4. The intelligent energy-saving control method for circular flotation machines based on an energy consumption model according to claim 3, characterized in that, The energy consumption field reconstruction engine, driven by the spatial distribution characteristics of the dynamic energy consumption trajectory line and the markers of state disturbance events, includes: The distribution density of dynamic energy consumption trajectory lines in a two-dimensional energy consumption field is analyzed, and grid cells of different sizes are used for initial partitioning in dense and sparse regions of dynamic energy consumption trajectory lines. Centered on each state disturbance event marker point, the deformation of the initial mesh element under local stretching or compression is calculated based on the type and magnitude of the state disturbance event. Based on the calculated deformation, the coordinates of all grid nodes in the initial grid cell are iteratively adjusted until the dynamic energy consumption trajectory segments of the state disturbance event marked by the deformed grid are evenly distributed in the grid, thus completing the generation of the reconstructed energy consumption field.

5. The intelligent energy-saving control method for circular flotation machines based on an energy consumption model according to claim 4, characterized in that, Based on the reconstructed energy consumption field, intelligent energy-saving control instructions for the circular flotation machine are generated, including: In the reconstructed energy consumption field, based on the distribution of the dynamic energy consumption trajectory line under the new nonlinear coordinates, the energy consumption cluster center and the energy consumption sparse channel are calculated. The energy consumption cluster center is defined by the region where the dynamic energy consumption trajectory line is densely concentrated, and the energy consumption sparse channel is defined by the path through which the dynamic energy consumption trajectory line is traversed at low density and without state disturbance event markers. The coordinate pairs of slurry concentration and motor power corresponding to the energy consumption sparse channel are extracted as a set of candidate energy-saving operating points. For each candidate energy-saving operating point in the candidate energy-saving operating point set, a pre-trained foam layer steady-state discrimination model is used for evaluation. The foam layer steady-state discrimination model takes the slurry concentration and motor power corresponding to the candidate energy-saving operating point as inputs and predicts the probability value that the foam layer height remains stable when the candidate energy-saving operating point is running. Candidate energy-saving operating points with probability values ​​exceeding the steady-state threshold are selected, and a list of target energy-saving operating points is generated. The list of target energy-saving operating points is input into the control command synthesizer, which maps each target energy-saving operating point to a motor speed control command and a slurry valve opening control command for the circular flotation machine. Based on the motor speed control command and the slurry valve opening control command, the operating status of the circular flotation machine is adjusted to achieve intelligent energy-saving control of the circular flotation machine; In the reconstructed energy consumption field, based on the distribution of the dynamic energy consumption trajectory line under the new nonlinear coordinates, the energy consumption clustering center and the energy consumption sparse channel are calculated, including: Under the nonlinear coordinates of the reconstructed energy consumption field, the dynamic energy consumption trajectory line is redrawn; The reconstructed energy consumption field is divided into multiple regular sub-regions, and the number of times the dynamic energy consumption trajectory line passes through each sub-region is counted to form a density distribution map; The center point of a sub-region whose density value exceeds the high density threshold is identified as the energy consumption accumulation center; Search outward from the energy consumption cluster center to find a continuous path that connects two energy consumption cluster centers and whose density values ​​of the sub-regions they pass through are all below the low density threshold. The continuous path is identified as an energy consumption sparse channel.

6. The intelligent energy-saving control method for circular flotation machines based on an energy consumption model according to claim 5, characterized in that, The extraction of coordinate pairs between slurry concentration and motor power from the energy-sparse channel, as a set of candidate energy-saving operating points, includes: Along each identified energy-sparse channel, multiple sampling locations are selected according to a preset equidistant sampling strategy; For each sampling point, based on its nonlinear coordinates in the reconstructed energy consumption field, the slurry concentration value and motor power value corresponding to the sampling point in the original two-dimensional energy consumption field are solved by using the inverse coordinate mapping relationship of the energy consumption field reconstruction engine. The coordinate pair consisting of the slurry concentration value and the motor power value is used as a candidate energy-saving operating point, and all candidate energy-saving operating points constitute the set of candidate energy-saving operating points.

7. The intelligent energy-saving control method for circular flotation machines based on an energy consumption model according to claim 6, characterized in that, The evaluation of each candidate energy-saving operating point in the candidate energy-saving operating point set using a pre-trained foam layer steady-state discrimination model includes: The foam layer steady-state discrimination model is a machine learning model trained based on historical operating data. Its input features include slurry concentration, motor power, and interaction features between slurry concentration and motor power. The slurry concentration value and motor power value of each candidate energy-saving operating point in the candidate energy-saving operating point set, as well as the interaction feature term calculated from the slurry concentration value and motor power value, are input into the foam layer steady-state discrimination model. The foam layer steady-state discrimination model outputs a foam layer steady-state probability value between zero and one. The foam layer steady-state probability value represents the predicted probability that the height of the foam layer in the circular flotation machine will remain within a preset stable range under the given combination of slurry concentration and motor power.

8. The intelligent energy-saving control method for circular flotation machines based on an energy consumption model according to claim 7, characterized in that, The process of filtering out candidate energy-saving operating points whose probability values ​​exceed the steady-state threshold generates a list of target energy-saving operating points, including: Set a steady-state probability threshold for the foam layer, and compare the probability value output by the steady-state discrimination model of the foam layer with the steady-state probability threshold of the foam layer; Retain all candidate energy-saving operating points whose probability values ​​are greater than or equal to the steady-state threshold; The remaining candidate energy-saving operating points are sorted in ascending order according to their corresponding motor power values ​​to generate an ordered list of target energy-saving operating points.

9. The intelligent energy-saving control method for circular flotation machines based on an energy consumption model according to claim 8, characterized in that, The control command synthesizer maps each target energy-saving operating point to a motor speed control command and a slurry valve opening control command for the circular flotation machine, including: Establish a mapping table in advance from motor power to motor speed, and a mapping table from slurry concentration to slurry valve opening; For each operating point in the target energy-saving operating point list, the mapping relationship table from motor power to motor speed is queried based on its motor power value to obtain the corresponding target motor speed value; For the same working point, the mapping relationship table between slurry concentration and slurry valve opening is consulted based on its slurry concentration value to obtain the corresponding target slurry valve opening value; The target motor speed value and the target slurry valve opening value are combined to form a set of motor speed control commands and slurry valve opening control commands.

10. The intelligent energy-saving control method for circular flotation machines based on an energy consumption model according to claim 9, characterized in that, Based on the motor speed control command and the slurry valve opening control command, adjust the operating status of the circular flotation machine, including: The motor speed control command is sent to the motor driver of the circular flotation machine, and the slurry valve opening control command is sent to the electric regulating valve of the slurry pipeline; The motor driver adjusts the motor to the target speed based on the received motor speed control command; The electric regulating valve adjusts the valve to the target opening degree according to the received slurry valve opening control command, thereby switching the circular flotation machine to the target energy-saving operating point.

Images