A method for optimizing energy consumption of periodic start-stop of a large-scale alkaline electrolyzer

Abstract

The present application relates to energy consumption optimization control technical field, specifically to a kind of large-scale alkaline electrolytic cell periodic start-stop energy consumption optimization control method, including the following steps, in the present application, preferred control sequence is obtained by calling genetic algorithm intercepting bit to perform exchange and constant disturbance operation, improve the global optimization convergence precision of discrete control instruction and reduce the integral value of overpotential loss associated with cold start-up power-up interval, by calling sparrow search algorithm to divide temperature difference group, extract vector superposition weight to update scalar, inhibit the optimization parameter divergence caused by bubble dense coverage effect associated with ohmic impedance nonlinear mutation of start-stop flow field, enhance the stability of AC power utilization efficiency under variable load wideband fluctuation condition, by writing numerical value to bottom storage unit to be converted into node trigger scale, compare system clock to issue pulse to gate output current to execute trajectory, forced input current tracks impedance bottom evolution curve, avoid invalid power conversion into excess waste heat.

Description

Technical Field

[0001] This invention relates to the field of energy consumption optimization and control technology, and in particular to a method for optimizing and controlling the energy consumption of a large alkaline electrolyzer during periodic start-up and shutdown. Background Technology

[0002] The field of energy consumption optimization and control technology specifically relies on operations research and nonlinear programming mathematical tools to perform power curve reshaping and energy loss suppression operations on ultra-high power industrial equipment. The field routinely covers core control models, including nonlinear autoregressive moving average models, Gaussian process regression surrogate models, and mixed-integer nonlinear programming optimization algorithms. The core logic is to transform the equipment's thermophysical coupling equations into a state-space matrix, set an energy consumption minimization objective function, apply physical boundary constraints, and then output a globally optimal discrete control command sequence matrix to avoid entering the inefficient polarization region during equipment start-up and shutdown, thereby reducing the overall integral power consumption of the system.

[0003] A method for optimizing energy consumption during the periodic start-stop cycle of a large alkaline electrolyzer involves constructing an explicit model predictive power scheduling scheme based on a multi-parameter quadratic programming algorithm. The scheme divides the continuous operating state space into polyhedral geometric regions and analyzes the affine control law mapping function for each region, replacing the time-consuming computation of conventional rolling optimization. The core objective of this scheme is to reduce the integral value of overpotential loss during the cold start-up power increase range and the standby / shutdown transition phase of the electrolyzer. Addressing the dense bubble coverage effect in the start-stop flow field and the nonlinear abrupt change in ohmic impedance, the scheduling scheme forces the input current to track the impedance valley evolution curve, avoiding the conversion of ineffective electrical energy into excess waste heat. This achieves a reduction in the total power consumption of the full-load start-stop cycle to a preset energy-saving baseline. It also constrains the rate of change of the internal temperature field gradient to be strictly below the critical threshold of thermal stress fatigue of core components, delaying the physical deformation of the microporous membrane. Ultimately, this ensures that the overall system maintains stable AC / DC power utilization efficiency within the target range under wide-frequency fluctuations in load, and mitigates the transient voltage drop impact induced by high-power step commands.

[0004] Traditional methods rely on multi-parameter quadratic programming algorithms to construct explicit model predictive power scheduling schemes, dividing the continuous operating state space into polyhedral geometric regions. However, the analytical affine control law mapping function operation exhibits significant limitations under actual wide-frequency fluctuation conditions. The fixed polyhedral geometric region division mechanism depends on the approximate linearization of the system's thermophysical coupling equations, failing to adaptively reconstruct the state space boundary to match the transient nonlinear physical abrupt changes in the flow field. During the cold start-up power increase range of the equipment, facing the rapid nonlinear jumps in ohmic impedance, the static affine control law cannot dynamically track the impedance valley evolution curve, causing the optimization trajectory to deviate from the global extremum and fall into a locally constrained state. When the transient temperature gradient on the diaphragm side approaches the critical threshold of thermal stress fatigue and the dense bubble coverage effect in the flow field intensifies, the inter-electrode impedance parameters exhibit a step-like discrete distribution. The system's preset explicit model predicts that the polyhedral partition cannot apply local coordinate reset operators and temperature penalty scalars, forcibly calling the hysteretic affine mapping function to continuously output the bias discrete control command sequence. This causes the AC / DC converter to follow the action with lag and induces the electrolytic cell to enter the inefficient polarization range. This not only forces a massive amount of input ineffective electrical energy to be converted into excess waste heat and increases the overall integral power consumption of the system, but also directly intensifies the thermophysical deformation process of core components and reduces the AC / DC power utilization efficiency of ultra-high power equipment during long-term operation. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing a method for optimizing energy consumption control during the periodic start-up and shutdown of large alkaline electrolyzers.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a method for optimizing and controlling the energy consumption of a large alkaline electrolytic cell during periodic start-up and shutdown, comprising the following steps: Step 1: Based on the input voltage of the bidirectional AC / DC converter and the initial temperature of the microporous diaphragm, extract the intersection mapping coordinates of the interelectrode impedance and the flow rate of the alkaline circulating pump, compare the coordinates with the pressure boundary array of the separator, remove out-of-bounds terms and splice the remaining terms to generate the thermoelectric state matrix. Step 2: Based on the thermoelectric state matrix, extract the transient potential derivative of the plates and the scalar of alkaline gas production rate, select the first-order array to calculate the crowding interval, call the genetic algorithm to truncate the bits and swap them with the constant perturbation, replace the vector and filter out out-of-bounds parameters to obtain the preferred control sequence; Step 3: Based on the preferred control sequence, extract the step power amplitude and gas diffusion layer coverage, calculate the upper and lower bound envelope overlap area variables, compare the area variables with the impedance jump array, select the items that meet the edge range, extract the horizontal axis time span value, and establish the communication cycle parameters. Step 4: Based on the communication cycle parameters, call the sparrow search algorithm to divide the temperature difference groups, extract the vector superimposed weight update scalar, calculate the difference superimposed tank temperature penalty term, determine the span trigger to reset the local scalar and filter the vector association duty cycle to obtain the control command matrix; Step 5: Based on the communication cycle parameters and the control command matrix, write the values ​​to the underlying storage unit and extract the parameters to convert them into node trigger scales. Compare the system clock with the scale to determine if they match and extract the message. Send pulses to the gate to execute the following action and output the current execution trajectory.

[0007] As a further embodiment of the present invention, the thermoelectric state matrix includes embedded node parameters of the intersection mapping coordinate grid, retained coordinate parameters, and system power step size limit vector; the preferred control sequence includes top-level undominated numerical parameters, fixed-word-length binary encoding sequence, and disturbance parameter vector; the communication cycle parameters include envelope area parameters, overshoot tolerance scalar, and geometric center abscissa scale; the control command matrix includes group update vector, embedded continuous co-position time span scalar of the stagnation determination array, and AC / DC converter duty cycle constant; and the current execution trajectory includes trigger control message, gate high / low level parameters, embedded phase shift conduction angle span parameters, and power step follower amplitude.

[0008] As a further aspect of the present invention, the specific steps for generating the thermoelectric state matrix are as follows: Based on the input voltage of the bidirectional AC / DC converter and the initial temperature of the microporous diaphragm, impedance parameters and circulating pump flow parameters are extracted, and the parameters are multiplied to construct a two-dimensional numerical array. The coordinates of the edge nodes of the overlapping region inside the array are extracted to obtain the intersection mapping coordinate grid. Based on the intersection mapping coordinate grid, the built-in node parameters of the grid are compared with the values ​​of the separator pressure boundary array. Coordinate parameters that do not meet the boundary conditions and are out of bounds are filtered out. The remaining coordinate parameters and the system power step size limit vector are spliced ​​together to generate the thermoelectric state matrix.

[0009] As a further aspect of the present invention, the specific steps for obtaining the preferred control sequence are as follows: Based on the thermoelectric state matrix, the electrode potential derivative and the reference scalar of alkaline gas production rate are extracted. The relative extreme value difference is calculated by comparing the scalar values. The difference is arranged to associate multidimensional parameter vectors and the multidimensional vector numerical dominance attribute is determined. The top-level undominated numerical parameters are separated to generate the first-level array. Based on the first-level array, the absolute distance between adjacent parameters is calculated and the distance values ​​are accumulated to give a crowding scalar. The scalar is sorted in descending order to extract the associated vector with a value greater than the threshold. The genetic algorithm operator is called to extract the bottom binary fragment of the vector, and the aligned bits are cross-transposed and superimposed with random constants to obtain the perturbation parameter vector. Based on the disturbance parameter vector, the original parent generation bottom-level parameters are extracted and fused with the parent generation parameters and the disturbance parameter vector to construct a candidate iteration parameter pool of the same dimension. The built-in parameters of the parameter pool are compared with the system power boundary threshold, parameters that exceed the system power boundary tolerance are eliminated, and the optimal control sequence is obtained.

[0010] As a further aspect of the present invention, the genetic algorithm extracts related vectors with a value greater than a threshold based on a descending scalar arrangement, calls an encoding operator to map the built-in numerical parameters of the related vectors with a value greater than the threshold to a fixed-length binary encoding sequence, extracts the total length scalar of the fixed-length binary encoding sequence, multiplies the total length scalar of the fixed-length binary encoding sequence by a preset truncation ratio constant to calculate the boundary position index, the truncation operator segments the fixed-length binary encoding sequence according to the boundary position index, extracts the bit data blocks after the boundary position index, and combines the bit data blocks to obtain the underlying binary fragment of the vector.

[0011] As a further aspect of the present invention, the specific steps for establishing the communication cycle parameters are as follows: Based on the preferred control sequence, the power step amplitude and diffusion layer coverage are extracted, the upper boundary curve and lower boundary line of the amplitude interval are set, the outer boundary line is spliced ​​to construct a closed geometric interval, the integral value of the overlapping area inside the interval is calculated, and the envelope area parameter is obtained. Based on the envelope area parameter, the parameter is compared with the impedance jump threshold matrix, low-order values ​​are removed and the boundary tolerance scalar is extracted. The scale of the geometric center of the retained scalar is mapped, the absolute difference of the time span between adjacent coordinates is calculated, and the communication cycle parameter is generated.

[0012] As a further aspect of the present invention, the specific steps for obtaining the control command matrix are as follows: Based on the communication cycle parameters, the parameters and the three-dimensional coordinates of the temperature difference boundary of the tank are merged. The sparrow search algorithm is called to decompose the numerical array into independent calculation sets. The velocity vector and position variables inside the set are extracted. The velocity and weight constant are multiplied and the coordinate difference is superimposed to obtain the group update vector. Based on the group update vector, the absolute difference between the individual vector and the reference boundary coordinates is extracted, the internal temperature gradient deviation penalty coefficient is matched and the penalty constant is accumulated to the end of the absolute difference scalar, and the continuous co-location time span is calculated by collecting numerical sequences to generate a stagnation determination array. Based on the stagnation determination array, the span variable is determined to exceed the clock tolerance boundary condition and the operator is triggered to change the local coordinate parameters. Divergent values ​​are compared and eliminated and low-cost row vectors are selected. The vectors are bound to the AC / DC converter duty cycle to establish the control command matrix.

[0013] As a further aspect of the present invention, the sparrow search algorithm constructs a numerical array based on the three-dimensional coordinates of the merged communication cycle parameters and the temperature difference boundary of the tank. The built-in parameters of the numerical array are mapped to the search population matrix. The total number of rows in the search population matrix is ​​extracted as a scalar. The first-order truncated index is calculated by multiplying the total number of rows in the search population matrix by a preset discoverer ratio constant. The top-level data block of the search population matrix is ​​disassembled according to the first-order truncated index to construct the discoverer calculation set. The last-order truncated index is calculated by multiplying the total number of rows in the search population matrix by a preset vigilant ratio constant. The bottom-level data block of the search population matrix is ​​disassembled according to the last-order truncated index to construct the vigilant calculation set. The remaining data block of the search population matrix is ​​extracted to construct the joiner calculation set. The discoverer calculation set, vigilant calculation set, and joiner calculation set are combined to divide the independent calculation sets.

[0014] As a further aspect of the present invention, the triggering operator iterates local coordinate parameters based on the stagnation determination array's built-in continuous co-position time span scalar exceeding the preset beat tolerance boundary condition. It generates an instruction to call the Levy flight mutation operator to perform coordinate translation on the stagnant individual. The mutation operator extracts a random step size vector with long-tail distribution probability characteristics, superimposes the random step size vector to the end of the current multidimensional spatial position coordinates of the stagnant individual, calculates and generates a large-scale random translation increment, adds the large-scale random translation increment to the initial local mapping coordinates to generate jump reset coordinates, and then directly overwrites and replaces the associated local coordinate parameters with the jump reset coordinates.

[0015] As a further aspect of the present invention, the specific steps for outputting the current execution trajectory are as follows: Based on the communication cycle parameters and the control instruction matrix, the values ​​are written to the underlying storage area and the parameters are converted to construct a trigger scalar. After collecting the hardware clock running scale, the absolute difference between the scale and the scalar is calculated. If the difference is zero, the instruction vector is intercepted and the trigger control message is obtained. Based on the trigger control message, the internal command scalar is parsed to convert high and low levels, the output level is sent to the converter gate port, and the built-in phase shift conduction angle span parameter is calculated. The angle parameter is bound to the duty cycle constant, and the power step follower amplitude is adjusted to obtain the current execution trajectory.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, a preferred control sequence is obtained by calling a genetic algorithm to extract bits and perform swapping and constant perturbation operations. This avoids the static multi-parameter quadratic programming analytical affine control law from falling into the local extremum trap in a high-dimensional nonlinear space, improves the global optimization convergence accuracy of discrete control commands, and reduces the integral value of overpotential loss associated with cold start power increase interval. In this invention, the sparrow search algorithm is used to divide temperature difference groups, extract vectors, superimpose weights, and update scalars. This suppresses the divergence of optimization parameters caused by the dense bubble coverage effect in the start-stop flow field and the nonlinear change in ohmic impedance, thereby enhancing the stability of AC / DC power utilization efficiency under variable load wide-frequency fluctuation conditions. In this invention, by writing numerical values ​​to the underlying storage unit and converting them into node trigger scales, the system clock sends pulses to the gate output current execution trajectory, forcing the input current to track the impedance valley evolution curve, thus avoiding the conversion of ineffective electrical energy into excess waste heat. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the main steps of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0019] Example 1 Please see Figure 1 This invention provides a technical solution: a method for optimizing and controlling the energy consumption of a large alkaline electrolyzer during periodic start-up and shutdown, comprising the following steps: Step 1: Based on the input voltage of the bidirectional AC / DC converter and the initial temperature of the microporous diaphragm, extract the intersection mapping coordinates of the interelectrode impedance and the flow rate of the alkaline circulating pump, compare the coordinates with the pressure boundary array of the separator, remove out-of-bounds terms and splice the remaining terms to generate the thermoelectric state matrix. Step 2: Based on the thermoelectric state matrix, extract the transient potential derivative of the plates and the scalar of alkaline gas production rate. Select the first and second arrays to calculate the crowding interval. Call the genetic algorithm to truncate the bits and swap them with the constant perturbation. After replacing the vector, filter out out-of-bounds parameters to obtain the optimal control sequence. Step 3: Based on the optimized control sequence, extract the step power amplitude and gas diffusion layer coverage, calculate the upper and lower bound envelope overlap area variables, compare the area variables with the impedance jump array, select the items that meet the edge range, extract the horizontal axis time span value, and establish the communication cycle parameters. Step 4: Based on the communication cycle parameters, the sparrow search algorithm is called to divide the temperature difference groups, extract the vector superimposed weight update scalar, calculate the difference superimposed tank temperature penalty term, determine the span trigger to reset the local scalar and filter the vector association duty cycle to obtain the control command matrix. Step 5: Based on the communication cycle parameters and control command matrix, write the values ​​to the underlying storage unit and extract the parameters to convert them into node trigger scales. Compare the system clock with the scale to determine if they match and extract the message. Send pulses to the gate to execute the following action and output the current execution trajectory.

[0020] The thermoelectric state matrix includes the intersection mapping coordinate grid built-in node parameters, retained coordinate parameters, and system power step size limit vector. The preferred control sequence includes top-level undominated numerical parameters, fixed word length binary encoding sequence, and disturbance parameter vector. The communication cycle parameters include envelope area parameters, overshoot tolerance scalar, and geometric center abscissa scale. The control command matrix includes group update vector, stagnation determination array built-in continuous co-position time span scalar, and AC / DC converter duty cycle constant. The current execution trajectory includes trigger control message, gate high and low level parameters, built-in phase shift conduction angle span parameter, and power step follower amplitude.

[0021] The specific steps for generating the thermoelectric state matrix are as follows: Based on the 650V input voltage of the bidirectional AC / DC converter and the initial temperature of 85°C on the microporous diaphragm side, a two-dimensional numerical array is constructed by multiplying the impedance parameter and the circulating pump flow parameter using the tensor Hadamard product operator. The Delaunay triangulation algorithm is then used to extract edges from the two-dimensional numerical array, generating an intersection mapping coordinate grid. Specifically, the Delaunay triangulation algorithm involves planning continuous storage blocks on the main control chip, pre-entering the circumscribed circle radius constant (2.55 mm) and the minimum interior angle limit parameter (15 degrees) into a read-only configuration file to establish preset edge extraction constraints, reading the spatial location point information built into the two-dimensional numerical array and loading it into the continuous storage block to form a discrete node group, and connecting adjacent points in the discrete node group. Generate an initial set of line segments, compare the polygon formed by the initial set of line segments with the minimum interior angle parameter of 15 degrees, adjust the connection relationship to avoid generating narrow and elongated geometric figures, construct a non-overlapping triangular mesh topological plane, calculate the geometric center coordinates and outer radius length of the circumcircle of all individual triangles contained in the non-overlapping triangular mesh topological plane, compare the outer radius length with the 2.55 mm constant built into the read-only configuration file, remove the outer individual triangles associated with the outer radius length greater than the 2.55 mm constant, merge the common boundary lines of the remaining adjacent individual triangles to construct a closed polygon outline, scan the outer boundary point sequence of the closed polygon outline, extract the horizontal and vertical dual-axis physical parameters of the outer boundary point sequence, and establish an intersection mapping coordinate grid; Based on the intersection-mapped coordinate grid, a spatial grid partitioning retrieval algorithm is used to compare and filter the built-in node parameters of the intersection-mapped coordinate grid with the values ​​of the separator pressure boundary array to generate a thermoelectric state matrix. Specifically, the spatial grid partitioning retrieval algorithm involves establishing a two-dimensional Cartesian coordinate system in memory, pre-inputting a spatial retrieval resolution side length of 0.05 mm via the terminal control panel and sending it to the system's main memory register to preset the retrieval grid constants, and then dividing the two-dimensional Cartesian coordinate system according to the 0.05 mm spatial retrieval resolution side length to generate multi-level uniform grid cells. The built-in node parameters of the intersection-mapped coordinate grid are mapped to the multi-level uniform grid cells and assigned index labels. The internal values ​​of the separator pressure boundary array are then read. The 1.2 MPa pressure limit data is used to locate the corresponding physical isolation boundary line in a two-dimensional Cartesian coordinate system. Multi-level uniform grid cells are scanned to compare the orientation of the physical isolation boundary line. Zero-value flag parameters are pre-written into the state memory to complete the boundary rule preset operation. It is determined that the multi-level uniform grid cells that cross the physical isolation boundary line carry zero-value flag parameters. All node coordinate parameters inside the grid cells that carry zero-value flag parameters are removed. The remaining coordinate parameters inside the grid cells that do not cross the physical isolation boundary line are extracted. The matrix column stacking operator is used to concatenate the remaining coordinate parameters with the pre-extracted system power step size limit vector along the vertical axis to close the outer boundary of the data sequence and generate a thermoelectric state matrix.

[0022] The specific steps for obtaining the preferred control sequence are as follows: Based on the thermoelectric state matrix, a row traversal reading operator is used to extract the electrode potential derivative and the alkali gas production rate reference scalar. A fast non-dominated sorting algorithm is employed to compare the scalar values, calculate the relative extreme value difference, arrange the differences, correlate the multi-dimensional parameter vectors, determine the dominance attribute of the multi-dimensional vector values, separate the top-level undominated numerical parameters, and generate a first-level array. Specifically, the fast non-dominated sorting algorithm allocates two free arrays of length 500 in the memory stack area, pre-deploys an extreme value limit constant of 0.05 via the underlying communication interface and writes it to a specific physical address in the register to complete the extreme value tolerance preset operation, reads the internal parameters of the multi-dimensional parameter vector row by row, calculates the absolute value of the relative extreme value difference between the electrode potential derivative and the alkali gas production rate reference scalar, traverses the 500 multi-dimensional parameter vectors, and compares the relative extreme value difference between any two parameter vectors. The absolute value and the register's built-in extreme value limit constant 0.05 are used to accumulate the number of times the current parameter vector is dominated by the remaining parameter vectors and assign the value to the domination counter variable. At the same time, the index labels of the current parameter vector dominating the remaining parameter vectors are recorded to construct a domination set linked list. The built-in value of the domination counter variable is scanned, and all parameter vectors whose domination counter variable is strictly equal to 0 are stripped. The stripped parameter vectors are assigned to a set with a level scalar of 1. For multiple parameter vectors within the set with a level scalar of 1, the built-in index labels of the domination set linked list are read. The value of the domination counter variable of the associated parameter vector is deducted one by one according to the index label. The counter deduction operation is continuously looped until the value of the domination counter variable returns to zero. The parameter vector that returns to zero for the first time is extracted and assigned to the top-level set area. Multiple parameters without domination values ​​in the top-level set area are concatenated to generate the first-level array. Based on the first-level hierarchical array, a congestion distance evaluation algorithm is used to calculate the cumulative absolute distance between adjacent parameters and assign it to the congestion scalar. The scalar values ​​are then sorted in descending order, and correlation vectors greater than a threshold are extracted. A congestion limit parameter of 1.5 is pre-input via a human-computer interface and saved to non-volatile memory to complete the threshold preset operation. Specifically, the congestion distance evaluation algorithm is executed as follows: the congestion scalar record array is initialized to 0; the parameters built into the first-level hierarchical array are sorted in ascending order according to the electrode potential derivative and the alkali gas production rate scalar; the congestion scalar values ​​of the first and last parameters are assigned an infinite constant; the parameter in the middle position is extracted, and the absolute distance between the physical scales of adjacent parameters is calculated. These two-dimensional absolute distance values ​​are added and assigned to the congestion scalar record array; a quicksort instruction is called to sort the congestion scalar record array in descending order; the sorting result is compared with the congestion limit parameter of 1.5 built into the non-volatile memory, and values ​​greater than 1 are extracted. .5. For the constant associated parameter vector, the multi-site crossover mutation algorithm is used to truncate the underlying binary fragment of the vector, align the bits for crossover, and superimpose random constants to obtain the perturbation parameter vector. Specifically, the multi-site crossover mutation algorithm maps continuous decimal parameters to 64-bit floating-point binary codes. The timestamp seed parameter 162345 is pre-injected into the random number generator to complete the preset operation of the random mechanism. The instruction specifies the 16th and 32nd bits of the binary code sequence as cutting nodes, truncates the underlying binary fragment between the 16th and 32nd bits, extracts the underlying binary fragments of two adjacent parameter vectors, performs corresponding bit data swapping, wakes up the random number generator to extract random constants in the range of 0.01 to 0.05, converts the random constants into binary sequences and superimposes them onto the end of the bit data swapping result, and reverse-parses the converted binary code into decimal physical parameters to obtain the perturbation parameter vector. Based on the perturbation parameter vector, a tensor splicing fusion operator is used to extract the original parent generation bottom-level parameters and fuse the parent generation parameters with the perturbation parameter vector to construct a candidate iteration parameter pool of the same dimension. A dynamic boundary constraint filtering algorithm is used to compare the built-in parameters in the parameter pool with the system power boundary threshold, eliminating parameters that exceed the system power boundary tolerance to obtain the optimal control sequence. Specifically, the dynamic boundary constraint filtering algorithm allocates a 1024-byte cache area in the system controller memory to store the built-in parameters of the candidate iteration parameter pool of the same dimension. The upper limit power of 3000 kW and the lower limit power of 500 kW data in the configuration file are read in advance and loaded into the hardware comparator logic gate to complete the system power boundary threshold preset operation. The candidate control parameters in the cache area are extracted one by one and transformed into tensor splicing fusion parameters. The hexadecimal machine code is fed into the input of the hardware comparator. The hardware comparator compares the input signal with the physical mapping digital quantities of the upper limit power of 3000 kW and the lower limit power of 500 kW in parallel. It monitors the output level signal of the hardware comparator status pin. The high-level signal is pre-bound to the rejection instruction through hardwire to complete the tolerance rule preset operation. When the hardware comparator status pin outputs a high-level signal, the underlying memory erase micro-instruction is activated. The erase triggers the high-level signal associated with the candidate control parameter data block. The low-level signal of the status pin is extracted and associated with the candidate control parameter. Multiple valid candidate control parameters that have not been erased are repackaged into a continuous data frame structure and stored in the main loop read register to generate the preferred control sequence.

[0023] The specific steps for establishing communication cycle parameters are as follows: Based on the optimized control sequence, the power step amplitude and diffusion layer coverage are extracted from the sequence using hexadecimal data unpacking instructions. A smoothness weighting coefficient of 0.85 is pre-input via the console keyboard and stored in a read-only memory block to complete the parameter preset process. A cubic spline interpolation algorithm is used to set the upper bound curve and lower bound polyline of the amplitude interval, and the outer polyline boundary is spliced ​​to construct a closed geometric interval. Specifically, the cubic spline interpolation algorithm allocates a three-dimensional coordinate array in the microprocessor memory, imports the power step amplitude and diffusion layer coverage into the three-dimensional coordinate array as interpolation nodes, constructs a tridiagonal linear equation system according to the smoothness weighting coefficient of 0.85, calls the chasing method to solve the tridiagonal linear equation system to calculate the second derivative values ​​of the multinomial nodes, connects the second derivative values ​​of the multinomial nodes to generate a smooth upper bound boundary line, uses linear interpolation to connect the bottom nodes to generate a lower bound polyline, extracts the coordinates of the first and last endpoints of the upper bound boundary line and the first and last endpoints of the lower bound polyline, and connects the first and last endpoints with straight lines to close the outer contour and construct a closed loop. For the geometric region, the adaptive Simpson integral algorithm is used to calculate the integral value of the overlapping area within the closed geometric region to obtain the envelope area parameter. The integral error limit parameter of 0.05% is written to the system configuration file in advance to complete the tolerance preset process. The adaptive Simpson integral algorithm is executed as follows: the closed geometric region is divided into 256 uniform initial sub-intervals along the horizontal axis. The horizontal axis coordinates of the midpoint of the initial sub-interval are read one by one, and the numerical integration operator is substituted to calculate the area of ​​the left geometric block and the cumulative scalar of the right geometric block respectively. The absolute difference of the cumulative scalar is compared with the integral error limit parameter of 0.05% built into the system configuration file. If the absolute difference is greater than the integral error limit parameter, the initial sub-interval is divided into two equal parts again and the cumulative calculation is performed iteratively. If the absolute difference is less than the integral error limit parameter, the division is stopped and the area value of the corresponding interval is written to the accumulation register. The 256 uniform initial sub-intervals are traversed and all values ​​in the accumulation register are merged to obtain the envelope area parameter. Based on the envelope area parameter, a matrix Boolean mask comparison algorithm is used to compare the envelope area parameter with the impedance transition threshold matrix, eliminating low-order values, extracting the boundary tolerance scalar, mapping and retaining the scalar's associated geometric center horizontal coordinate scale. A transition threshold of 75 milliohms is pre-deployed via a host computer and fixed to the comparator reference level pin to complete the boundary pre-setting process. Specifically, the matrix Boolean mask comparison algorithm reads a 128x128 impedance transition threshold matrix from memory, extracts the parameters within the impedance transition threshold matrix and the envelope area parameter item by item, and multiplies them using a hardware multiplier to obtain the physical amplification scalar. The physical amplification scalar is then compared in parallel with the 75 milliohm transition threshold built into the comparator reference level pin. For physical amplification scalars less than 75 milliohms, a zero-value flag is written to the memory physical address to perform low-order value elimination. For physical amplification scalars greater than 75 milliohms, a one-value flag is written to the corresponding address to extract the value. The system retrieves the row and column index coordinates of a two-dimensional array containing a single-value flag, and uses an affine transformation operator to convert these coordinates into the horizontal coordinate scale of the geometric center in physical space, generating a set of retained timestamps. A one-dimensional differential operator is then used to calculate the absolute difference in time span between adjacent coordinates within the retained timestamp set, generating communication cycle parameters. A 10-millisecond constant base clock period is pre-written to the system timer to complete the baseline preset process. Specifically, the one-dimensional differential operator sorts the horizontal coordinate scales within the retained timestamp set in ascending order according to time sequence and loads them into a one-dimensional buffer queue. The instruction arithmetic logic unit sequentially reads the current and subsequent scale scalars from the one-dimensional buffer queue, performs decimal subtraction to obtain the absolute value of the distance between them, multiplies this absolute value by the 10-millisecond base clock period constant to convert it into a physical time span scalar, and merges all physical time span scalars to construct a discrete time interval sequence, generating the communication cycle parameters.

[0024] The specific steps to obtain the control command matrix are as follows: Based on the communication cycle parameter, a tensor dimensionality reduction and splicing operator is used to merge the communication cycle parameter and the three-dimensional coordinates of the tank temperature difference boundary to generate an initial mapping coordinate matrix. A sparrow search algorithm is then used to decompose the initial mapping coordinate matrix into a numerical array, dividing it into independent computation sets. The velocity vectors and position variables within each set are extracted, multiplied by the velocity and weight constants, and then superimposed with coordinate differences to obtain the group update vector. Specifically, the sparrow search algorithm allocates a continuous data storage block with a population size constant of 100 in the system memory. The discoverer proportion constant of 0.2 and the vigilant proportion constant of 0.1 are pre-distributed via the host computer communication bus and stored in the system read-only register to complete the proportion constant preset operation. The initial mapping coordinate matrix is ​​read and loaded into the continuous data storage block. The total number of rows in the initial mapping coordinate matrix is ​​calculated and multiplied by the discoverer proportion constant stored in the read-only register. The truncation position index scalar is calculated using the number 0.2 and the vigilant proportional constant 0.1. Following the truncation position index scalar, the initial mapped coordinate matrix is ​​horizontally divided into the discoverer matrix, the vigilant matrix, and the joiner matrix to construct independent computation sets. For multiple computation units within the independent computation sets, the velocity vector associated with the current iteration step size and the spatial three-dimensional position variables are extracted. The inertia weight constant 0.65 is extracted from the environmental parameter library. The hardware multiplier is instructed to perform the multiplication operation of the velocity vector and the inertia weight constant 0.65 to generate a weighted velocity scalar. The current spatial three-dimensional position variables and the global extreme value three-dimensional position variables are extracted and fed into the hardware subtractor to perform vector subtraction to calculate the physical coordinate difference. The hardware adder is instructed to perform the superposition operation of the weighted velocity scalar and the physical coordinate difference. The superimposed value is extracted to cover the original spatial three-dimensional position variables and generate a group update vector. Based on the group update vector, a dynamic exterior point penalty function model is used to extract the absolute difference between the individual vector and the reference boundary coordinates, match the internal temperature gradient deviation penalty coefficient, and accumulate the penalty constant to the end of the absolute difference scalar to generate a penalized deviation scalar. An Euclidean distance time series aggregation algorithm is used to calculate the continuous co-location time span of the penalized deviation scalar through the aggregated numerical sequence to generate a stagnation judgment array. Specifically, the dynamic exterior point penalty function model reads the current individual vector containing instantaneous operation parameters and the corresponding safety threshold of the reference boundary coordinates, inputs both to the absolute value subtractor to calculate the physical deviation value, and pre-sets the penalty parameters by inputting a temperature deviation amplification factor of 1.5 and a temperature gradient limit of 5 degrees Celsius via keyboard input on the control panel and writing them to the non-volatile storage area. The current temperature gradient scalar measured by the temperature sensor is read, and if it is determined that the temperature gradient scalar is greater than the temperature gradient limit of 5 degrees Celsius, a hardware multiplier is activated to exponentiate the temperature deviation amplification factor of 1.5 and the excess value. The system calculates a penalty constant scalar, instructs the arithmetic logic unit to add the penalty constant scalar to the end of the physical deviation numerical data segment to obtain the penalized deviation scalar, and executes the Euclidean distance time series aggregation algorithm. Specifically, it establishes a timestamp queue in the system cache, pre-writes a sampling period of 50 milliseconds and a spatial overlap judgment threshold of 0.01 millimeters to the system timer to complete the preset operation of the judgment condition, reads the physical space three-dimensional coordinates associated with the penalized deviation scalar at the current time and the physical space three-dimensional coordinates at the previous time according to the sampling period of 50 milliseconds, and sends them to the distance calculation core module to calculate the absolute value of the Euclidean distance between the two sets of physical space three-dimensional coordinates, compares the absolute value of the Euclidean distance with the spatial overlap judgment threshold of 0.01 millimeters, triggers the internal accumulator counter to perform an increment operation for values ​​less than 0.01 millimeters to accumulate the continuous same position time span, and triggers a clear pulse to reset the value of the internal accumulator counter to zero for values ​​greater than 0.01 millimeters, merges all single-vector associated accumulator counter output scalars to construct a one-dimensional array, and generates a stagnation judgment array. Based on the stagnation determination array, the Levy flight mutation operator is used to determine if the span variable exceeds the beat tolerance boundary condition, triggering the operator to iterate local coordinate parameters and generate a reset coordinate array. A nonlinear mapping algorithm is used to compare and eliminate divergent values, filter low-cost row vectors, bind vectors to the AC / DC converter duty cycle operation, and establish a control command matrix. Specifically, the Levy flight mutation operator is executed by pre-writing a continuous stagnation limit of 15 constants to the tolerance register via the communication bus to complete the tolerance boundary preset operation. It then iterates through the internal accumulator counter of the stagnation determination array, outputting a scalar. If the output scalar is greater than the continuous stagnation limit of 15 constants, a random walk generator is triggered to extract a long-distance jump step size scalar following a heavy-tailed probability distribution. Local coordinate parameters associated with the stagnation state are extracted, and the long-distance jump step size scalar and local coordinate parameters are added to generate a large-scale translation reset coordinate. The memory controller is then instructed to use the large-scale translation reset coordinate. The process involves replacing the original local coordinate parameters with a standard overlay to generate a reset coordinate array. The execution of the nonlinear mapping algorithm specifically involves opening a floating-point filtering channel within the processor, pre-setting the maximum duty cycle parameter (0.9) and the minimum duty cycle parameter (0.1) in the duty cycle limiter to complete the physical boundary preset operation, substituting the internal parameters of the reset coordinate array into the system preset cost function calculation module to output multiple target cost scalars, calling a quicksort instruction to perform ascending sorting on the multiple target cost scalars, extracting the row vectors associated with the first 10 low-order cost scalars from the sorted queue, inputting the low-order cost row vectors to the nonlinear activation module to calculate continuous output probability values, multiplying the continuous output probability values ​​by the maximum duty cycle parameter (0.9) to calculate the physical drive signal amplitude, comparing the physical drive signal amplitude with the minimum duty cycle parameter (0.1), and forcing values ​​below 0.1 to be assigned a constant of 0.1. The corrected physical drive signal amplitude is then encapsulated to construct a two-dimensional data table structure, establishing a control instruction matrix.

[0025] The specific steps for the output current execution trajectory are as follows: Based on the communication cycle parameters and control instruction matrix, the direct memory access bus write mechanism is used to write values ​​to the underlying storage area of ​​the above two-dimensional array to package and construct the data block to be sent. A clock tick conversion mapping operator is used to convert parameters of the data block to be sent to construct a trigger scalar and generate a hardware trigger scalar. Specifically, the clock tick conversion mapping operator is executed by first writing a division constant of 8 to the internal clock divider register through the underlying system application program interface to complete the clock reference preset operation. The system main frequency crystal oscillator frequency of 100 MHz is read and divided by the division constant of 8 to calculate the physical time scalar of 80 nanoseconds for a single tick. The built-in time span value of the data block to be sent is extracted and divided by the physical time scalar of 80 nanoseconds for a single tick to convert the physical time parameter into a discrete integer tick constant, which is written to a specific offset address in the register to form a hardware trigger scalar. A high-frequency hardware timer interrupt comparison algorithm is used to collect the hardware trigger scalar, and the absolute value between the hardware clock running scale and the scale is calculated. The process involves determining the difference and resetting it to zero, extracting the instruction vector, obtaining the trigger control message, and executing a high-frequency hardware timer interrupt comparison algorithm. Specifically, this involves pre-registering the starting address pointer of the zero-value matching interrupt service function in the system interrupt vector table to complete the interrupt mechanism preset operation. A 32-bit free-running upward counter is started to continuously accumulate and generate a hardware clock running scale. The hardware comparison logic gate is instructed to synchronously latch the hardware clock running scale and the hardware trigger scalar every clock cycle and send them to the subtraction logic circuit to perform binary subtraction and output the absolute difference signal. The level of the absolute difference signal is monitored. When all data bits of the absolute difference signal are at zero level, the hardware comparison logic gate outputs a single pulse high-level signal to the kernel interrupt control pin, triggering the microprocessor main core to push the stack, save the current context, and jump to the zero-value matching interrupt service function. The register is read from the main loop to extract the row vector of the matching time-related control instruction matrix, establish an independent data frame structure, and generate the trigger control message. Based on the trigger control message, a pulse-width modulation (PWM) digital logic parsing operator is used to parse the trigger control message, convert the internal instruction scalar to high and low levels, and output the level to the converter gate port to generate a physical drive signal. Specifically, the PWM digital logic parsing operator opens a direct output digital channel in the microcontroller unit, pre-writes the push-pull output mode identifier code 01 to the general-purpose input / output port configuration register to complete the pin state preset operation, reads the binary valid bits of the built-in instruction scalar in the trigger control message bit by bit, triggers the set register for data with a binary valid bit of 1, injects a continuous voltage of 3.3 volts into the direct output digital channel, and triggers the reset register for data with a binary valid bit of 0, pulling the direct output digital channel low to a 0-volt reference low level, maintaining the port level state until the end of the next clock cycle to form an edge step pulse. This edge step pulse is transmitted to the AC / DC converter's insulated-gate bipolar transistor gate physical interface via an optocoupler isolation circuit to form a physical drive signal. A phase-shift space vector control algorithm is used to calculate the physical drive signal, binding the built-in phase-shift conduction angle span parameter with the angle parameter and duty cycle constant to control the power step follower amplitude operation to obtain the current. The execution trajectory and phase-shift space vector control algorithm specifically involve: pre-inputting a dead-time parameter of 2 microseconds via a human-machine interface serial terminal and storing it in the dead-time generator to complete the safety timing preset operation; reading the trigger time misalignment difference of the parallel physical drive signals of the AC / DC converter's multi-arms, multiplying it by the AC power grid's fundamental frequency angular velocity of 314 radians per second constant, and converting it to obtain the multi-arm trigger phase lag angle parameter, i.e., the built-in phase-shift conduction angle span parameter; extracting the pre-stored duty cycle constant from the memory area; and instructing the hardware multiply-accumulate unit to perform a dot multiplication operation on the built-in phase-shift conduction angle span parameter and the duty cycle constant to generate... The amplitude scaling factor is used to read the DC bus rated reference power scalar of 1500 kW, multiply it by the amplitude scaling factor to calculate the discrete value of the target power jump within the current command cycle, and send the discrete value of the target power jump to the underlying digital-to-analog converter to be converted into a continuous stepped analog reference voltage. The input to the converter hardware current loop comparator performs closed-loop negative feedback regulation of the current loop based on the continuous stepped analog reference voltage, and collects the continuous current amplitude data points of the Hall sensor at the output of the hardware current loop. The continuous current amplitude data points are spliced ​​along the time axis sequence to construct an evolution polyline and obtain the current execution trajectory.

[0026] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for optimizing and controlling the energy consumption of a large-scale alkaline electrolytic cell during periodic start-up and shutdown, characterized in that, Includes the following steps: Step 1: Based on the input voltage of the bidirectional AC / DC converter and the initial temperature of the microporous diaphragm, extract the intersection mapping coordinates of the interelectrode impedance and the flow rate of the alkaline circulating pump, compare the coordinates with the pressure boundary array of the separator, remove out-of-bounds terms and splice the remaining terms to generate the thermoelectric state matrix. Step 2: Based on the thermoelectric state matrix, extract the transient potential derivative of the plates and the scalar of alkaline gas production rate, select the first-order array to calculate the crowding interval, call the genetic algorithm to truncate the bits and swap them with the constant perturbation, replace the vector and filter out out-of-bounds parameters to obtain the preferred control sequence; Step 3: Based on the preferred control sequence, extract the step power amplitude and gas diffusion layer coverage, calculate the upper and lower bound envelope overlap area variables, compare the area variables with the impedance jump array, select the items that meet the edge range, extract the horizontal axis time span value, and establish the communication cycle parameters. Step 4: Based on the communication cycle parameters, call the sparrow search algorithm to divide the temperature difference groups, extract the vector superimposed weight update scalar, calculate the difference superimposed tank temperature penalty term, determine the span trigger to reset the local scalar and filter the vector association duty cycle to obtain the control command matrix; Step 5: Based on the communication cycle parameters and the control command matrix, write the values ​​to the underlying storage unit and extract the parameters to convert them into node trigger scales. Compare the system clock with the scale to determine if they match and extract the message. Send pulses to the gate to execute the following action and output the current execution trajectory.

2. The energy consumption optimization control method for periodic start-up and shutdown of a large-scale alkaline electrolytic cell according to claim 1, characterized in that, The thermoelectric state matrix includes built-in node parameters of the intersection mapping coordinate grid, retained coordinate parameters, and system power step size limit vector. The preferred control sequence includes top-level undominated numerical parameters, fixed-word-length binary encoding sequence, and disturbance parameter vector. The communication cycle parameters include envelope area parameters, overshoot tolerance scalar, and geometric center abscissa scale. The control command matrix includes group update vector, built-in continuous co-position time span scalar of the stagnation determination array, and AC / DC converter duty cycle constant. The current execution trajectory includes trigger control message, gate high and low level parameters, built-in phase shift conduction angle span parameter, and power step follower amplitude.

3. The energy consumption optimization control method for periodic start-up and shutdown of a large-scale alkaline electrolytic cell according to claim 1, characterized in that, The specific steps for generating the thermoelectric state matrix are as follows: Based on the input voltage of the bidirectional AC / DC converter and the initial temperature of the microporous diaphragm, impedance parameters and circulating pump flow parameters are extracted, and the parameters are multiplied to construct a two-dimensional numerical array. The coordinates of the edge nodes of the overlapping region inside the array are extracted to obtain the intersection mapping coordinate grid. Based on the intersection mapping coordinate grid, the built-in node parameters of the grid are compared with the values ​​of the separator pressure boundary array. Coordinate parameters that do not meet the boundary conditions and are out of bounds are filtered out. The remaining coordinate parameters and the system power step size limit vector are spliced ​​together to generate the thermoelectric state matrix.

4. The energy consumption optimization control method for periodic start-up and shutdown of a large-scale alkaline electrolytic cell according to claim 1, characterized in that, The specific steps for obtaining the preferred control sequence are as follows: Based on the thermoelectric state matrix, the electrode potential derivative and the reference scalar of alkaline gas production rate are extracted. The relative extreme value difference is calculated by comparing the scalar values. The difference is arranged to associate multidimensional parameter vectors and the multidimensional vector numerical dominance attribute is determined. The top-level undominated numerical parameters are separated to generate the first-level array. Based on the first-level array, the absolute distance between adjacent parameters is calculated and the distance values ​​are accumulated to give a crowding scalar. The scalar is sorted in descending order to extract the associated vector with a value greater than the threshold. The genetic algorithm operator is called to extract the bottom binary fragment of the vector, and the aligned bits are cross-transposed and superimposed with random constants to obtain the perturbation parameter vector. Based on the disturbance parameter vector, the original parent generation bottom-level parameters are extracted and fused with the parent generation parameters and the disturbance parameter vector to construct a candidate iteration parameter pool of the same dimension. The built-in parameters of the parameter pool are compared with the system power boundary threshold, parameters that exceed the system power boundary tolerance are eliminated, and the optimal control sequence is obtained.

5. The energy consumption optimization control method for periodic start-up and shutdown of a large-scale alkaline electrolytic cell according to claim 1, characterized in that, The genetic algorithm extracts related vectors with a value greater than a threshold based on a descending scalar arrangement. It calls an encoding operator to map the built-in numerical parameters of the related vectors with a value greater than the threshold to a fixed-length binary encoding sequence. It extracts the total length scalar of the fixed-length binary encoding sequence, multiplies the total length scalar of the fixed-length binary encoding sequence by a preset truncation ratio constant to calculate the boundary position index, and the truncation operator segments the fixed-length binary encoding sequence according to the boundary position index. It extracts the bit data blocks after the boundary position index and combines the bit data blocks to obtain the underlying binary fragment of the vector.

6. The method for optimizing and controlling the energy consumption of a large-scale alkaline electrolytic cell during periodic start-up and shutdown according to claim 1, characterized in that, The specific steps for establishing the communication cycle parameters are as follows: Based on the preferred control sequence, the power step amplitude and diffusion layer coverage are extracted, the upper boundary curve and lower boundary line of the amplitude interval are set, the outer boundary line is spliced ​​to construct a closed geometric interval, the integral value of the overlapping area inside the interval is calculated, and the envelope area parameter is obtained. Based on the envelope area parameter, the parameter is compared with the impedance jump threshold matrix, low-order values ​​are removed and the boundary tolerance scalar is extracted. The scale of the geometric center of the retained scalar is mapped, the absolute difference of the time span between adjacent coordinates is calculated, and the communication cycle parameter is generated.

7. The method for optimizing and controlling the energy consumption of a large-scale alkaline electrolytic cell during periodic start-up and shutdown according to claim 1, characterized in that, The specific steps to obtain the control command matrix are as follows: Based on the communication cycle parameters, the parameters and the three-dimensional coordinates of the temperature difference boundary of the tank are merged. The sparrow search algorithm is called to decompose the numerical array into independent calculation sets. The velocity vector and position variables inside the set are extracted. The velocity and weight constant are multiplied and the coordinate difference is superimposed to obtain the group update vector. Based on the group update vector, the absolute difference between the individual vector and the reference boundary coordinates is extracted, the internal temperature gradient deviation penalty coefficient is matched and the penalty constant is accumulated to the end of the absolute difference scalar, and the continuous co-location time span is calculated by collecting numerical sequences to generate a stagnation determination array. Based on the stagnation determination array, the span variable is determined to exceed the clock tolerance boundary condition and the operator is triggered to change the local coordinate parameters. Divergent values ​​are compared and eliminated and low-cost row vectors are selected. The vectors are bound to the AC / DC converter duty cycle to establish the control command matrix.

8. The method for optimizing and controlling the energy consumption of a large-scale alkaline electrolytic cell during periodic start-up and shutdown according to claim 1, characterized in that, The sparrow search algorithm constructs a numerical array based on the merging communication cycle parameters and the three-dimensional coordinates of the tank temperature difference boundary. The built-in parameters of the numerical array are mapped to a search population matrix. A scalar of the total number of rows in the search population matrix is ​​extracted. This scalar is multiplied by a preset discoverer ratio constant to calculate the first-order truncation index. Following this first-order truncation index, the top-level data block of the search population matrix is ​​disassembled to construct a discoverer calculation set. A scalar of the total number of rows in the search population matrix is ​​multiplied by a preset vigilant ratio constant to calculate the last-order truncation index. Following this last-order truncation index, the bottom-level data block of the search population matrix is ​​disassembled to construct a vigilant calculation set. The remaining data block of the search population matrix is ​​extracted to construct a joiner calculation set. The discoverer calculation set, vigilant calculation set, and joiner calculation set are then combined to create independent calculation sets.

9. The method for optimizing and controlling the energy consumption of a large-scale alkaline electrolytic cell during periodic start-up and shutdown according to claim 7, characterized in that, The triggering operator iterates local coordinate parameters based on the stagnation determination array's built-in continuous co-position time span scalar exceeding the preset beat tolerance boundary condition. It generates an instruction to call the Levy flight mutation operator to perform coordinate translation on the stagnant individual. The mutation operator extracts a random step vector with long-tail distribution probability characteristics and superimposes the random step vector to the end of the stagnant individual's current multidimensional spatial position coordinates. It calculates and generates a large-scale random translation increment. The random translation increment is added to the initial local mapping coordinates to generate jump reset coordinates. Then, the jump reset coordinates directly cover and replace the associated local coordinate parameters.

10. The method for optimizing and controlling the energy consumption of a large-scale alkaline electrolytic cell during periodic start-up and shutdown according to claim 1, characterized in that, The specific steps for outputting the current execution trajectory are as follows: Based on the communication cycle parameters and the control instruction matrix, the values ​​are written to the underlying storage area and the parameters are converted to construct a trigger scalar. After collecting the hardware clock running scale, the absolute difference between the scale and the scalar is calculated. If the difference is determined to be zero, the instruction vector is intercepted and the trigger control message is obtained. Based on the trigger control message, the internal command scalar is parsed to convert high and low levels, the output level is sent to the converter gate port, and the built-in phase shift conduction angle span parameter is calculated. The angle parameter is bound to the duty cycle constant, and the power step follower amplitude is adjusted to obtain the current execution trajectory.

Images