A photovoltaic water pumping intelligent scheduling and power optimization method

Abstract

The present application belongs to the technical field of data processing, and discloses a photovoltaic water pumping intelligent scheduling and power optimization method, steps comprising: S1 collecting historical data of the target agricultural irrigation area, and preprocessing the data; S2 constructing a mode coupling characteristic mining and scheduling situation map of photovoltaic output and water demand time sequence; S3 through scheduling situation guidance and multi-constraint self-healing improved ant colony optimization algorithm, embedding the scheduling situation map into the dynamic visibility calculation, gear shift probability, pheromone update and multi-constraint self-healing repair links of ant colony optimization, forming a multi-objective scheduling solution method suitable for photovoltaic water pumping tasks. The photovoltaic output and irrigation water demand are respectively decomposed into different frequency scale intrinsic mode components, which can identify the nonlinear misplacement relationship of photovoltaic output peak and irrigation water demand peak on different time scales, solving the problem that traditional moving average or single-scale linear correlation analysis is difficult to describe multi-scale supply and demand misplacement.

Description

Technical Field

[0001] This invention belongs to the field of data processing technology, and in particular relates to a method for intelligent scheduling and power optimization of photovoltaic water lifting. Background Technology

[0002] Photovoltaic water lifting technology combines solar power generation with water pumping, providing a clean and sustainable irrigation water supply method for remote irrigation areas with weak power grid coverage. In recent years, with the continuous decline in the cost of photovoltaic modules and the increasing demand for agricultural irrigation electrification, photovoltaic water lifting systems have been widely promoted globally, becoming an important way to solve the dual challenges of agricultural energy consumption and water resource allocation. However, photovoltaic water lifting systems face a series of complex technical challenges in actual operation, such as: 1. Existing technologies mostly use moving average, single correlation coefficient or fixed period load matching methods to analyze the relationship between photovoltaic output and irrigation water demand. They can only capture the superficial trend of time series or the linear correlation under a single scale, and it is difficult to reveal the coupling strength and supply and demand peak misalignment law of photovoltaic daily cycle, cloud shadow short-term disturbance and irrigation multi-peak demand at different frequency scales.

[0003] 2. Existing scheduling methods typically input photovoltaic power, water demand, reservoir status, and electricity price directly into the optimization model as independent boundary conditions or constraints. They lack feature engineering methods to integrate these multidimensional heterogeneous information into a unified scheduling prior, resulting in a lack of directional guidance in the early stages of the search, which affects both convergence efficiency and solution quality.

[0004] 3. Existing ant colony optimization or other heuristic methods for visibility design in photovoltaic water pumping scheduling are mostly based on fixed rules, such as only considering the size of photovoltaic power or the level of electricity price. They cannot dynamically adjust the attractiveness of each level according to the comprehensive situation of the current scheduling period, and it is also difficult to simultaneously take into account photovoltaic absorption, electricity price response and level switching smoothness.

[0005] 4. When dealing with strong temporally coupled constraints such as upper and lower limits of water volume in reservoirs, minimum start and stop time of water pumps, and power variation limits, existing methods usually only perform feasibility assessment during the path generation stage and directly discard infeasible solutions. They lack the ability to repair infeasible paths after the fact, resulting in a large amount of search computation being wasted. Especially when the constraints are strict, feasible solutions are scarce or even the search fails.

[0006] The aforementioned technical challenges are intertwined, making it extremely difficult to achieve efficient, economical, and reliable scheduling operations. Summary of the Invention

[0007] The purpose of this invention is to provide a method for intelligent scheduling and power optimization of photovoltaic water pumping, in order to solve the problems of low optimization efficiency, poor solution quality, and difficulty in balancing economy and reliability in existing photovoltaic water pumping scheduling methods, which have deficiencies in supply and demand coupling analysis, multi-source information fusion, heuristic search guidance, and constraint violation handling.

[0008] The technical solution adopted in this invention is a method for intelligent scheduling and power optimization of photovoltaic water lifting, the steps of which include: Step S1: Collect historical data of the target agricultural irrigation area and preprocess the data; Step S2: Construct a modal coupling feature mining and scheduling situation map of photovoltaic power output and water demand time series; Step S3 involves improving the ant colony optimization algorithm by guiding the scheduling situation and using multi-constraint self-healing. The scheduling situation map is embedded into the dynamic visibility calculation, gear transfer probability, pheromone update and multi-constraint self-healing repair links of ant colony optimization, forming a multi-objective scheduling solution method suitable for photovoltaic water lifting tasks.

[0009] Furthermore, the data collected in S1 includes historical measured available photovoltaic power output, historical irrigation water demand, and the available photovoltaic power, irrigation water demand, real-time water volume in the reservoir, and time-of-use electricity price pump performance parameters predicted for the current scheduling cycle. The above data are used to construct a historical available photovoltaic power output sequence in chronological order. Historical irrigation water demand sequence and the available photovoltaic power sequence predicted for the current scheduling cycle. Irrigation water demand sequence Time-of-use electricity pricing series ; The specific steps of the preprocessing are as follows: S11, unify all multi-data samples into the same scheduling period; S12, correct abnormal data in the historical photovoltaic available output sequence and historical irrigation water demand sequence after unified scheduling time axis, including: 1) data missing correction; 2) abnormal mutation data correction.

[0010] Furthermore, the specific steps of S2 are as follows: S21, Adaptive variational mode decomposition is performed on the preprocessed historical photovoltaic available power output sequence and historical irrigation water demand sequence to obtain photovoltaic intrinsic mode components and water demand intrinsic mode components, and the center frequency of each photovoltaic intrinsic mode component and each water demand intrinsic mode component is recorded. S22, pair the photovoltaic intrinsic mode components with the water demand intrinsic mode components, and comprehensively measure the information sharing strength and center frequency proximity between each pair of modes to calculate the mutual information coupling coefficient between the photovoltaic intrinsic mode components and the water demand intrinsic mode components; S23, Selecting mutual information coupling coefficients greater than the selection threshold. The mode pairs are taken as the dominant resonant mode pairs. The dominant resonant mode pairs are reconstructed by weighting according to the coupling strength to obtain the dominant photovoltaic resonant sequence. Dominant water demand resonance sequence Harmony and light-water resonance components; S24, dynamically time-bending aligned the dominant photovoltaic resonance sequence and the dominant water demand resonance sequence, in order to Using the reference sequence, find the minimum matching path between the two sequences and establish an alignment mapping function. Divide the time index offset of each scheduling period by the number of scheduling periods per day to obtain the optical-water matching offset. S25, the predicted available photovoltaic power, predicted irrigation water demand, photovoltaic-water matching offset, reservoir water status and time-of-use electricity price in the current scheduling cycle are normalized to [0,1] by minimum-maximum, and the scheduling status scalar of each scheduling period is obtained by weighted nonlinear fusion, and then arranged in chronological order to form a scheduling status diagram.

[0011] Furthermore, the specific steps of S22 are as follows: S22a, respectively, the first The photovoltaic intrinsic mode components, the first The modal values ​​of the intrinsic mode components of water demand over all scheduling periods constitute the photovoltaic intrinsic mode vector. Water demand intrinsic mode vector The photovoltaic intrinsic mode vector and the water demand intrinsic mode vector are adaptively binned, and the mutual information coupling coefficient is calculated based on the joint distribution after binning, as shown in the following formula: In the formula, Indicates the first The photovoltaic intrinsic mode component and the first The mutual information coupling coefficient between the intrinsic modal components of water demand. The coupling coefficient is used to calculate the stability constant. This represents the frequency bandwidth tolerance parameter. Represents the natural exponential function; This represents the absolute value operator. This represents the mutual information calculation function. Indicates the first Information entropy of photovoltaic intrinsic mode vectors Indicates the first Information entropy of the intrinsic mode vector of water demand Indicates the first The center frequency of each photovoltaic intrinsic mode component Indicates the first The center frequency of the intrinsic modal components of water demand; S22b, Calculate the frequency domain penalty weight based on the center frequency difference between the photovoltaic intrinsic mode components and the water demand intrinsic mode components. ; S22c, iterate through all photovoltaic intrinsic mode components and all water demand intrinsic mode components, calculate all mutual information coupling coefficients, and combine them to obtain the mutual information coupling coefficient matrix. .

[0012] Furthermore, the specific steps of S23 are as follows: S23a, Screening for mutual information coupling coefficients greater than the screening threshold The mode pairs are considered as dominant resonance mode pairs, forming the dominant resonance mode pair set. and to The photovoltaic intrinsic mode components and water demand intrinsic mode components in the data are standardized with zero mean and unit variance, respectively, to obtain... Photovoltaic normalized intrinsic mode components at time and water demand normalized intrinsic mode components ; S23b, calculate the normalized coupling weights based on the mutual information coupling coefficients of each dominant resonance mode pair, and reconstruct the dominant photovoltaic resonance sequences according to the normalized coupling weights. Resonance sequence with dominant water demand ; S23c, by weighted fusion of the normalized intrinsic mode components of photovoltaics and the normalized intrinsic mode components of water demand, yields the photovoltaic-water resonance component. .

[0013] Furthermore, the specific steps of S3 are as follows: S31 uses the pump operating speed sequence throughout the entire scheduling cycle as the ant's search path, searches for the Pareto approximate optimal solution set of the scheduling scheme, and constructs the scheduling path for each ant. ,in, Indicates the first The scheduling path constructed by an ant Indicates the first Only ants in the first The pump operating speed selected for the specified time period; S32, calculate the dynamic visibility of the scheduling situation guidance, for the first... In the nth iteration Only ants in the first The first scheduling period For each pump operating speed, the dynamic scheduling status scalar is combined with the ratio of pump power to rated power to calculate the status adaptation term; the electricity price sensitivity index is used as the power of the ratio of benchmark electricity price and time-of-use electricity price to calculate the electricity price adjustment term for each scheduling period; the absolute value of the difference between the current speed power and the selected speed power in the previous scheduling period and the negative value of the ratio of the power change tolerance parameter are used as the exponent of the natural constant e to calculate the speed switching smoothing term; finally, the dynamic visibility is calculated by combining the status adaptation term, power matching term, electricity price suppression term, and speed smoothing term. S33, calculate the transfer probability of each pump's operating gear based on the current pheromone concentration, dynamic visibility, and gear feasibility constraints, and generate a complete scheduling path for each time period according to the probability. S34. After the candidate scheduling path is generated, perform multi-constraint self-healing repair, locate the critical period that causes the constraint violation along the time direction, and select the alternative position with the lowest correction cost from the candidate repair positions to transform the infeasible scheduling path into a feasible candidate scheduling path. S35 performs multi-objective evaluation and Pareto ranking on all candidate scheduling paths, only uses non-dominated solutions for pheromone enhancement, and improves the pheromone contribution of sparse region solutions by crowding distance. S36, when the ant colony search reaches the maximum number of iterations. If the Pareto frontier changes are below the frontier change threshold for multiple consecutive rounds, output a set of non-dominated candidate scheduling schemes. Each candidate scheduling scheme includes the pump operation speed sequence, corresponding power allocation, water volume trajectory of the reservoir, and multi-objective evaluation results for the entire scheduling cycle. Finally, the execution scheme is selected from the final candidate scheduling scheme set based on actual operation preferences.

[0014] Furthermore, the specific steps of S31 are as follows: S31a, read the... Pump power and predicted reservoir water volume for time-scheduled routes and the Predicted water volume in reservoirs for different time periods , will the Water flow rate corresponding to the pump operating speed during the time period and scheduling interval Multiply by the product to obtain the water extraction volume; then use the predicted irrigation water demand. The outflow of water is deducted, and the constraint violation of the scheduling path is calculated through state deduction. All constraint violations are recorded. Specifically: if the predicted water volume of the reservoir is lower than the minimum allowable water volume of the reservoir, a water supply shortage or a violation of the lower limit of the water level will occur. If the predicted water volume of the reservoir is higher than the maximum allowable water volume of the reservoir, an overflow risk or a violation of the upper limit of the water level will occur. S31b, the target values ​​of operating cost, water supply deficit, curtailment loss and gear switching are concatenated into a four-dimensional vector to obtain the multi-objective evaluation vector of each scheduling path; S31c specifies the design capacity of the water storage tank and the minimum continuous operating time of the water pump. Minimum continuous downtime of water pump and maximum gear change The constraint is used as a hard constraint to verify each path on a time-by-time basis, and to filter out scheduling paths that are not feasible to actually execute.

[0015] Furthermore, the specific steps of S33 are as follows: S33a, Establish the pheromone matrix corresponding to the scheduling period and the pump operating speed; S33b, according to the Only one ant has been identified as the previous one. Based on the pump operating speed, the current predicted reservoir water volume, the minimum continuous operating time of the pump, the minimum continuous shutdown time of the pump, and the maximum power change, the first... The set of feasible gears for each scheduling period ; S33c, combining pheromone concentration, dynamic visibility, and feasible gear set, calculates the... In the nth iteration Only ants in the first Select the first scheduling period The formula for the transition probability of each pump's operating speed is as follows: In the formula, Indicates the first In the nth iteration Only ants in the first Select the first scheduling period The probability of switching between different operating speeds of a water pump; This indicates the pheromone influence index. The dynamic visibility impact index indicates the impact of visibility. This indicates the index for traversing the pump's operating speed. This indicates the feasibility indicator function for gear selection. Indicates the first In the nth iteration The scheduling period and the first The pheromone concentration corresponding to each pump operating speed Indicates the first In the nth iteration Only ants in the first Select the first scheduling period Dynamic visibility at each pump operating speed; Indicates the first In the nth iteration Only ants in the first Select the first scheduling period When the water pump is in its operating gear, the first Dynamic visibility at each pump operating speed; Indicates the first A set of feasible gears for each scheduling period; S33d, perform roulette wheel sampling on the transition probabilities corresponding to each feasible pump operating position to determine the first... In the nth iteration Only ants in the first Pump operating speed during each scheduling period and set the water pump to the correct operating speed. Add to the current candidate scheduling path; after determining the pump operating speed, update the temporary reservoir water volume according to the water volume status extrapolation method, and update the pump operating status, continuous running time or continuous shutdown time according to the selected pump operating speed, and then proceed to the next step. For each scheduling period, recalculate the dynamic scheduling situation scalar, dynamic visibility, feasible gear set, and transition probability, traversing from the 1st to the th... After the scheduling period, the first... Complete candidate scheduling path for each ant .

[0016] Furthermore, the specific steps of S34 are as follows: S34a: Perform forward simulation on the complete candidate scheduling path. Starting from the first scheduling period, calculate the pump power, water lifting volume, actual available irrigation water volume, water supply deficit, reservoir water volume, pump start / stop status, continuous running time, and continuous downtime for each period, and record all constraint violation periods; and record the scheduling period in which the earliest constraint violation occurs during the forward simulation as... ,exist Candidate repair time periods were identified within the previous backtracking window, and the backtracking window length was set. ; S34b, for each candidate repair period From the set of selectable speeds of the water pump The pump operating speeds that meet the constraints of pump power boundary, minimum continuous operating time, minimum continuous downtime, and maximum power change are selected to form a candidate repair speed set. ; S34c: Based on the power change of the candidate pump operating position relative to the original pump operating position, and the degree of water volume exceeding the limit during the original constraint violation period after the re-forward simulation, the correction cost is calculated through the penalty coefficient. All candidate repair periods and their candidate repair positions are traversed, and the candidate repair period and pump operating position with the minimum correction cost are selected. S34d, the candidate scheduling path is in the... The selected repair period corresponds to the original water pump operating speed, which is then replaced with the repair speed. And from the first The forward simulation will be re-executed at the start of each scheduling period. If there are still constraint violations, the earliest constraint violation period will be re-determined and the backtracking location, candidate repair level selection, and minimum correction cost selection will continue.

[0017] Furthermore, the specific steps of S35 are as follows: S35a, calculate the... In each iteration, all ants generate multi-objective evaluation vectors for candidate scheduling paths after multi-constraint self-healing repair. Pareto non-dominated sorting is then performed to determine if any candidate scheduling path dominates another candidate scheduling path. Candidate scheduling paths that are not dominated by other candidate scheduling paths form a non-dominated solution set. ; S35b, calculate the normalized congestion distance for candidate scheduling paths in the non-dominated solution set, and then... The normalized crowding distance in the target space is denoted as ; S35c performs minimum-maximum normalization on the target values ​​of operating cost, water supply deficit, curtailment loss, and shift switching, respectively, to obtain four normalized target values ​​for candidate scheduling paths; then, the candidate scheduling path is calculated based on the weighted sum of the four normalized target values. Corresponding basic pheromone increment ; S35d involves the evaporation and enhancement of pheromones. The pheromone renewal method is represented as follows: in, Indicates the first In the nth iteration Time period and the The pheromone concentration corresponding to each pump operating speed Indicates the first Pheromones evaporation rate in each iteration. Indicates the crowding distance amplification coefficient. This indicates the path / gear indicator function. Indicates the candidate scheduling path at the th The pump operating speed corresponding to the time period.

[0018] The beneficial effects of this invention are: 1. This invention can decompose photovoltaic power output and irrigation water demand into intrinsic mode components of different frequency scales, and can identify the nonlinear misalignment relationship between peak photovoltaic power output and peak irrigation water demand at different time scales, thus solving the problem that traditional moving average or single-scale linear correlation analysis is difficult to characterize multi-scale supply and demand misalignment.

[0019] 2. This invention can normalize and weight-integrate the predicted available photovoltaic power, predicted irrigation water demand, photovoltaic-water matching offset, reservoir water status, and time-of-use electricity price into a scheduling status scalar for each scheduling period. This transforms the originally discrete multi-dimensional operational information into a single continuous signal of water lifting suitability, providing an intuitive and physically meaningful task prior for the optimization algorithm.

[0020] 3. By embedding the scheduling status scalar into the dynamic visibility calculation of ant colony optimization, the ants can perceive whether the photovoltaic power is sufficient, whether the water use is urgent, whether the reservoir is short of water, and whether the electricity price is peak when selecting the water pump operation level. Combined with power matching, electricity price suppression, and level switching smoothing, the photovoltaic-water coupling characteristics are transmitted from the data analysis layer to the optimization decision layer, realizing data-driven heuristic search.

[0021] 4. The multi-constraint self-healing repair mechanism designed in this invention transforms infeasible scheduling paths into feasible candidate paths, avoiding the waste of search resources caused by traditional optimization methods that directly discard infeasible solutions. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a basic time-series data display diagram of the present invention, in which (a) is historical photovoltaic power output, (b) is historical irrigation water demand, (c) is prediction comparison, and (d) is time-of-use electricity price.

[0024] Figure 2 The results of the adaptive variational mode decomposition analysis of the present invention are as follows: (a) are the five intrinsic mode components of the historical photovoltaic available power output sequence decomposition, and (b) are the four intrinsic mode components of the historical irrigation water demand sequence decomposition.

[0025] Figure 3This invention provides an analysis of dynamic visibility and pheromone concentration distribution.

[0026] Figure 4 This is a diagram illustrating the scheduling path for ant colony generation according to the present invention.

[0027] Figure 5 This is a diagram showing the water volume trajectory and power allocation of the reservoir under the scheduling path of the present invention, wherein (a) is a curve of water volume change and safety range, and (b) is a stacked display diagram of photovoltaic available power, water pump power consumption, grid-purchased power and curtailed power. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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.

[0029] An embodiment of the present invention provides a method for intelligent scheduling and power optimization of photovoltaic water lifting, the steps of which include: Step S1: Collect seven types of multivariate data from the target agricultural irrigation area over a past time period (set according to specific needs, such as 1 year, 1 month, etc.), including historical measured available photovoltaic power output, historical irrigation water demand, and the predicted available photovoltaic power, irrigation water demand, and real-time water volume in the reservoir during the current scheduling cycle (the water volume at the start time is recorded as...). The minimum allowable water volume is recorded as The maximum allowable water volume is recorded as The performance parameters of the water pump at different times of the electricity price were analyzed, and a sequence was constructed in chronological order as follows: Historical Available Photovoltaic Output Sequence Historical irrigation water demand sequence and the available photovoltaic power sequence predicted for the current scheduling cycle. Irrigation water demand sequence Time-of-use electricity pricing series The data samples were preprocessed. (Time series data visualization diagram) Figure 1 As shown, (a) represents historical photovoltaic power output, (b) represents historical irrigation water demand, (c) represents predicted comparison, and (d) represents time-of-use electricity pricing. The pump speed performance parameters include a set of selectable speeds denoted as... Water pump power The water pump operating speed corresponding to the water flow rate The specific steps of preprocessing are as follows: S11 unifies all multi-data samples to the same scheduling period. If the sampling interval of a certain original data source is less than the scheduling interval, then average, integrate, or count the maximum value of multiple sampling points within the same scheduling period. If the sampling interval of a certain original data source is greater than the scheduling interval, then linear interpolation, keeping the previous value, or similar daily curve completion are used to convert to a unified scheduling time axis.

[0030] S12, corrects abnormal data in the historical available photovoltaic power output sequence and historical irrigation water demand sequence after the unified scheduling time axis, to ensure that the photovoltaic supply side and irrigation load side are compared at the same time granularity. Specifically: 1) Data missing correction: when the length of the missing segment does not exceed the missing threshold. When the missing data exceeds a threshold, linear interpolation between adjacent valid time periods is used for padding. When missing data is found, the average of historical similar days within the same season, weather type, or irrigation task type is used to fill in the missing data. The missing threshold is [not specified in the original text]. This represents the maximum number of consecutive missing time periods that can be filled using local interpolation, preferably 3 to 6.

[0031] 2) Correction of abnormal mutation data: For values ​​in the historical photovoltaic available output sequence after unified scheduling time axis that are greater than 0 at night or whose photovoltaic power changes exceed the allowable ramp-up range of the equipment's rated power, and for values ​​in the historical irrigation water demand sequence after unified scheduling time axis that exceed the pump station's maximum water supply capacity and have no corresponding manual scheduling record, median filtering or quantile clipping is performed.

[0032] Step S2 involves constructing a modal coupling feature mining and scheduling status map of photovoltaic power output and water demand time series to reflect the scheduling status of water lifting suitability during each scheduling period. The specific steps are as follows: S21. Adaptive variational mode decomposition is performed on the preprocessed historical available photovoltaic power output sequence and historical irrigation water demand sequence, respectively. The complex time series is decomposed into several intrinsic mode components at different frequency scales, namely photovoltaic intrinsic mode components and water demand intrinsic mode components. The center frequency of each photovoltaic intrinsic mode component and each water demand intrinsic mode component is recorded (it can be directly output by the variational mode decomposition algorithm, or it can be obtained by performing a fast Fourier transform on the intrinsic mode components and selecting the main peak frequency with the largest amplitude). The decomposition results are as follows: Figure 2 As shown, (a) represents the five intrinsic mode components of the historical available photovoltaic power output sequence decomposition, and (b) represents the four intrinsic mode components of the historical irrigation water demand sequence decomposition. The center frequency of each photovoltaic intrinsic mode component is denoted as... The center frequency of the intrinsic modal component of water demand is denoted as The decomposition relation can be expressed as: In the formula, This represents the historical available photovoltaic power output sequence. Represents the historical irrigation water demand sequence. Indicates the first The photovoltaic intrinsic mode component in the th... Modal values ​​for a given time period This represents the photovoltaic residual component, which is the low-energy perturbation in the historical available photovoltaic power output sequence that is not fully explained by the photovoltaic intrinsic mode components; This represents the number of photovoltaic eigenmodes obtained from the final decomposition of the historical available photovoltaic power output sequence. Indicates the first The intrinsic modal component of water demand in the th Modal values ​​for a given time period are used to characterize the changes in historical irrigation water demand sequences at the corresponding frequency scale; This represents the residual component of water demand, which is the low-energy perturbation in the historical irrigation water demand sequence that is not fully explained by the intrinsic modal component of water demand. This represents the total number of intrinsic water demand modes obtained from the final decomposition of the historical irrigation water demand sequence. Indicates the first The center frequency of each photovoltaic intrinsic mode component is used to describe the main fluctuation period of that photovoltaic intrinsic mode component; Indicates the first The center frequency of each intrinsic mode component of water demand is used to describe the main fluctuation period of that intrinsic mode component of water demand.

[0033] S22, pair the photovoltaic intrinsic mode components with the water demand intrinsic mode components, and comprehensively measure the information sharing strength and center frequency proximity between each pair of modes to calculate the mutual information coupling coefficient between the photovoltaic intrinsic mode components and the water demand intrinsic mode components. This coefficient is used to measure the degree of nonlinear dependence between the two. The specific steps are as follows: S22a, respectively, the first The photovoltaic intrinsic mode components, the first The modal values ​​of the intrinsic mode components of water demand over all scheduling periods constitute the photovoltaic intrinsic mode vector. Water demand intrinsic mode vector The photovoltaic intrinsic mode vector and the water demand intrinsic mode vector are adaptively binned, and the mutual information coupling coefficient is calculated based on the joint distribution after binning, as shown in the following formula: In the formula, Indicates the first The photovoltaic intrinsic mode component and the first The mutual information coupling coefficient between the intrinsic modal components of water demand is used to quantify the comprehensive matching degree between the two in terms of information correlation and frequency proximity. The larger the value, the stronger the time-series coupling of light and water. This represents the stability constant used in calculating the coupling coefficients, used to prevent the denominator from being zero; it can be taken as... ; This represents the frequency bandwidth tolerance parameter, used to control the penalty strength of the center frequency difference on the mutual information coupling coefficient, and can be taken as 0.05~0.15; Represents the natural exponential function; The absolute value operator is used to calculate the distance between two center frequencies; This represents the mutual information calculation function. Indicates the first The information entropy of a photovoltaic intrinsic mode vector (the specific calculation method belongs to the existing technology). Indicates the first Information entropy of the intrinsic mode vector of water demand Indicates the first The center frequency of each photovoltaic intrinsic mode component Indicates the first The center frequency of the intrinsic modal component of water demand.

[0034] S22b, Calculate the frequency domain penalty weight based on the center frequency difference between the photovoltaic intrinsic mode components and the water demand intrinsic mode components. The formula is as follows: In the formula, , These are the center frequencies of the two intrinsic mode components, respectively. The reference frequency bandwidth is used to control the penalty strength of the center frequency difference on the mutual information coupling coefficient, and can be taken as 0.05~0.15.

[0035] If the information sharing strength of the two intrinsic mode components obtained in S22a is high, but the difference in their center frequencies is too large, it indicates that the two may only have an accidental correspondence in local numerical changes and are not suitable as a supply and demand coupling relationship under the same time scale.

[0036] S22c, iterate through all photovoltaic intrinsic mode components and all water demand intrinsic mode components, calculate all mutual information coupling coefficients, and combine them to obtain the mutual information coupling coefficient matrix. ,in, The dimension is , This represents the number of photovoltaic eigenmodes obtained from the final decomposition of the historical available photovoltaic power output sequence. This represents the total number of water demand eigenmodes obtained from the final decomposition of the historical irrigation water demand sequence. Each element in the mutual information coupling coefficient matrix corresponds to the coupling strength between a photovoltaic eigenmode component and a water demand eigenmode component.

[0037] S23, based on the mutual information coupling coefficient calculated in S22c, filters out those with mutual information coupling coefficients greater than a screening threshold. The mode pairs are taken as the dominant resonant mode pairs. The dominant resonant mode pairs are weighted and reconstructed according to the coupling strength to obtain the dominant photovoltaic resonant sequence, the dominant water demand resonant sequence, and the photovoltaic-water resonant component. This process removes weakly coupled modes and residual noise, highlighting the supply and demand coupling frequency bands that have a practical contribution to scheduling. The specific steps are as follows: S23a, Based on the mutual information coupling coefficient matrix, filter out those with mutual information coupling coefficients greater than the screening threshold. The mode pairs are considered as dominant resonance mode pairs, forming the dominant resonance mode pair set. and to The photovoltaic intrinsic mode components and water demand intrinsic mode components in the data are standardized with zero mean and unit variance, respectively, to obtain... Photovoltaic normalized intrinsic mode components at time and water demand normalized intrinsic mode components The filtering threshold Take the first element of all elements in the coupling coefficient matrix percentile, The preferred range is 60~80.

[0038] S23b, calculate the normalized coupling weights based on the mutual information coupling coefficients of each dominant resonance mode pair, and reconstruct the dominant photovoltaic resonance sequences according to the normalized coupling weights. Resonance sequence with dominant water demand These are used to characterize the changes in available photovoltaic power output and irrigation water demand in the key coupling frequency band, respectively. The calculation formulas are as follows: In the formula, Indicates the first The photovoltaic intrinsic mode component and the first Normalized coupling weights corresponding to the intrinsic modal components of water demand. Indicates the first The photovoltaic intrinsic mode component and the first The mutual information coupling coefficient between the intrinsic modal components of water demand. Indicates the first The photovoltaic intrinsic mode component and the first The mutual information coupling coefficient between the intrinsic modal components of water demand. This represents the index for traversing photovoltaic modes in the set of dominant resonant modes. This represents the index of the dominant resonance mode's traversal of water demand modes in the set. express The photovoltaic normalized intrinsic mode components at time t, express Normalized intrinsic mode components of water demand at any given time. This represents the number of photovoltaic eigenmodes obtained from the final decomposition of the historical available photovoltaic power output sequence. This represents the total number of intrinsic water demand modes obtained from the final decomposition of the historical irrigation water demand sequence.

[0039] S23c fuses the normalized intrinsic mode components of photovoltaics and the normalized intrinsic mode components of water demand to obtain the photovoltaic-water resonance component. The formula is as follows: in, The light-water resonance component is represented in the first... The time period values ​​are used to characterize the intensity of the coordinated change between available photovoltaic power output and irrigation water demand in the dominant coupling frequency band. The higher the value, the more significant the number. The more consistent the changes in photovoltaic power output and irrigation water demand are on key frequency scales around a given time period, the better. The lower the value, the weaker the synergistic relationship between light and water during that period or the existence of a supply-demand mismatch. express The photovoltaic normalized intrinsic mode components at time t, express Normalized intrinsic modal components of water demand at any given time.

[0040] S24, for the dominant photovoltaic resonance sequence Resonance sequence with dominant water demand Perform dynamic time warp (DTW) alignment to Using the reference sequence, find the minimum matching path between the two sequences and establish an alignment mapping function. The time index offset of each scheduling period is divided by the number of scheduling periods per day to obtain the optical-water matching offset, as shown in the following formula: In the formula, It refers to the first The solar-water matching offset during a given time period is used to characterize the degree of misalignment between available photovoltaic power output and irrigation water demand in time. The smaller the value, the closer the supply and demand peaks are, and the larger the value, the more obvious the misalignment between the supply and demand peaks. This indicates the number of scheduling periods per day, used to normalize the time offset, and the preferred value is 96. It is an alignment mapping function, representing the alignment with the first... A time period index of a reference sequence that matches a scheduling period is used to quantify the degree of misalignment between supply and demand peaks; Indicates the first The time index distance between a time period and its matching time period is used to directly measure the degree of misalignment between supply and demand peaks.

[0041] S25, the predicted available photovoltaic power, predicted irrigation water demand, photovoltaic-water matching offset, reservoir water status, and time-of-use electricity price within the current scheduling cycle are normalized to [0,1] using a minimum-maximum normalization method. Then, a weighted nonlinear fusion is used to obtain the scheduling status scalar for each scheduling period. These scalars are then arranged in chronological order to form a scheduling status diagram, providing prior information for subsequent ant colony optimization scheduling. The formula is as follows: In the formula, Indicates the first Normalization of time periods can be achieved using photovoltaic power. Indicates the first Normalized irrigation water demand for different time periods Indicates the first Normalized reservoir water volume over a given period of time Indicates the first Normalized time-of-use electricity pricing for different time periods This represents the logical mapping function (i.e., the Sigmoid function). Indicates the first The scheduling status scalar for the time period. , , , , These represent the weights for normalized available photovoltaic power, normalized irrigation water demand, photovoltaic-water matching offset, normalized reservoir water volume, and normalized time-of-use electricity price, respectively. The preferred values ​​are as follows: , , , , ,satisfy .

[0042] Step S3 involves improving the ant colony optimization algorithm through scheduling situation guidance and multi-constraint self-healing. The scheduling situation diagram is embedded into the dynamic visibility calculation, gear shift probability, pheromone update, and multi-constraint self-healing repair stages of ant colony optimization, forming a multi-objective scheduling solution method suitable for photovoltaic water lifting tasks. This method achieves a balance between operating costs, water supply reliability, photovoltaic absorption capacity, and equipment operational stability. The dynamic visibility and pheromone concentration distribution diagram is shown below. Figure 3 As shown, the specific steps are as follows: S31 uses the pump operating speed sequence throughout the entire scheduling cycle as the ant's search path to search for the Pareto approximate optimal solution set of the scheduling scheme. Each ant represents a complete scheduling scheme search, and the scheduling path constructed by each ant is... ,in, Indicates the first The scheduling path constructed by the ant includes the pump operating speed for each scheduling period within the entire scheduling cycle. Indicates the first Only ants in the first The pump operating speed selected for the time period is used to determine the first... The scheduling path for ant colony generation is as follows: pump power and water flow rate during different time periods. Figure 4 As shown, the specific search steps are as follows: S31a, read the... Pump power and predicted reservoir water volume for time-scheduled routes and the Predicted water volume in reservoirs for different time periods , will the Water flow rate corresponding to the pump operating speed during the time period and scheduling interval Multiply by the product to obtain the water extraction volume; then use the predicted irrigation water demand obtained from S1. By deducting the outflow, the system calculates and records constraints that violate the scheduling path through state deduction, providing a triggering basis for subsequent multi-objective evaluation (water supply deficit objective) and multi-constraint self-healing repair. Specifically: if the predicted water volume in the reservoir is lower than the minimum allowable water volume, a water supply deficit or lower water level violation occurs; if the predicted water volume in the reservoir is higher than the maximum allowable water volume, an overflow risk or upper water level violation occurs.

[0043] The water output Based on forecasted irrigation water demand Specific: If ,but ;otherwise The shortfall will be included in the water supply deficit. ,in Indicates the first The water flow rate corresponding to the pump's operating speed during a given time period.

[0044] S31b, calculate the multi-objective evaluation vector for each scheduling path, using the following formula: in, Indicates the first The multi-objective evaluation vector corresponding to the ant scheduling path is used for Pareto ranking and candidate scheme comparison; This represents the target operating cost, which is obtained by summing up the electricity purchase costs under the time-of-use pricing. The smaller the value, the better the economic efficiency. This represents the target value for water supply deficit, which is obtained by accumulating the amount of water that does not meet the irrigation water demand in each scheduling period. The smaller the value, the higher the reliability of water supply. The target value for curtailment loss is obtained by summing up the photovoltaic power that is not used by water pumps and cannot be connected to the grid or stored during each scheduling period. The smaller the value, the stronger the photovoltaic absorption capacity. This represents the target value for gear switching, which is obtained by summing the changes in pump operating gear or pump power during adjacent scheduling periods. The smaller the value, the more stable the pump operation.

[0045] S31c specifies the design capacity of the water storage tank and the minimum continuous operating time of the water pump. Minimum continuous downtime of water pump and maximum gear change Constraints (pump parameters are set according to specific requirements) are used as hard constraints. Each path is verified on a time-by-time basis to filter out scheduling paths that are not feasible. The specific constraints are: (a) (b) Continuous running time (c) Continuous downtime (d) .

[0046] Only in (a) is there a single time period exceeding the limit and deviation amount. If any constraint (b), (c), or (d) is violated or the water volume exceeds the limit, it will be marked as infeasible. If the violation still occurs after repair, it will be marked as infeasible.

[0047] S32, Calculate the dynamic visibility of the dispatch situation guidance. The specific steps are as follows: S32a, for the first In the nth iteration Only ants in the first The first selectable scheduling period For each pump operating speed, the dynamic scheduling status scalar is combined with the ratio of pump power to rated power to calculate the status adaptation term, as shown in the following formula: In the formula, Indicates the first In the nth iteration Only ants in the first Select the first scheduling period The degree of adaptation between the operating speed of each water pump and the current dynamic scheduling situation. Indicates the first In the nth iteration Only ants in the first Select the first scheduling period The degree of matching between the operating speed of each water pump and the current power. Indicates the rated power of the water pump. Indicates the first In the nth iteration Only ants in the first The scheduling status scalar for each scheduling period. Indicates the rated power of the water pump. This indicates the power matching tolerance parameter, which is taken as 15% to 25% of the rated power of the water pump; Indicates the first The water pump power corresponding to each water pump operating speed. Indicates the first The normalization of time periods can be achieved using photovoltaic power.

[0048] when At that time, it can be considered that the first The operating speed of each water pump is close to the predicted available photovoltaic power; when At that time, the first Each pump's operating speed may require a significant amount of electricity purchased from the grid; when At that time, the first The operating speed of the water pump may result in a significant amount of photovoltaic power not being absorbed by the water pump.

[0049] S32b calculates the electricity price adjustment term and the tier switching smoothing term using the following formula: In the formula, Indicates the first The degree to which time-of-use electricity pricing affects the selection of pump operating speeds during each scheduling period. This represents the benchmark electricity price, which is the average of all time-of-use electricity prices within the current dispatch cycle; Indicates the first Time-of-use electricity price for each dispatch period, This represents the electricity price sensitivity index, used to control the intensity of the impact of time-of-use electricity prices on dynamic visibility; in this embodiment, it is preferably 1. Indicates the first In the nth iteration Only ants in the first Select the first scheduling period The smoothness of the pump operation speed relative to the previous scheduling period when each pump operates at a certain speed. This represents the power variation tolerance parameter, which is preferably 30% of the pump's rated power. Indicates the first The water pump power corresponding to each water pump operating speed; Indicates the first The water pump power corresponding to each water pump operating speed. Indicates the first In the nth iteration Only ants in the first The pump operating speed selected for the time period is used to determine the first... Pump power and water flow rate during the specified time period.

[0050] when When the electricity price adjustment item is greater than 1, the priority of tier selection is increased during periods of low electricity prices; when When the electricity price adjustment item is less than 1, it suppresses unnecessary high-power water pumping during periods of high electricity prices.

[0051] The smaller the change in pump power, the closer the gear shifting smoothness term is to 1; the larger the change in pump power, the smaller the gear shifting smoothness term is, thereby reducing the probability of selecting the appropriate pump operating gear for frequent starts and stops or drastic gear changes.

[0052] S32c, combining situational adaptation, power matching, electricity price suppression, and speed smoothing, calculates dynamic visibility to measure the heuristic priority of the pump's operating speed under the current dispatching situation. A higher value indicates a more suitable selection, as shown in the formula below: In the formula, Indicates the first In the nth iteration Only ants in the first Select the first scheduling period The dynamic visibility of each pump's operating speed is the highest that indicates the pump's operating speed is more suitable to be selected under the current reservoir water level, available photovoltaic power, irrigation water demand, time-of-use electricity price, and the previous pump's operating speed. The index represents the impact of the scheduling situation, which is used to control the intensity of the impact of the dynamic scheduling situation on the dynamic visibility. In this embodiment, it is preferably 2.

[0053] It should be noted that dynamic visibility does not simply encourage selecting the maximum water pump operating level when the photovoltaic power is high. Instead, it simultaneously limits the deviation between water pump power and photovoltaic power, time-of-use electricity price, and power changes in adjacent scheduling periods. Based on this, it can avoid blindly increasing the water pump operating level during high photovoltaic periods, which would cause the water volume in the reservoir to exceed the upper limit. It can also avoid only considering low electricity prices while ignoring the urgency of water supply.

[0054] S33. Based on the current pheromone concentration, dynamic visibility, and gear feasibility constraints, calculate the transition probability of each pump's operating gear, and generate a complete scheduling path for each time period according to the probability. The specific steps are as follows: S33a, establish the pheromone matrix corresponding to the scheduling period and the pump operating speed. In the... In the nth iteration, the 1st The scheduling period and the first The pheromone concentration corresponding to each pump operating speed is denoted as The higher the pheromone concentration, the more likely it is to occur during the historical search process in the [number]th [period]. Select the first scheduling period The more excellent candidate scheduling paths there are for each pump's operating speed, the better.

[0055] In this invention, before the first iteration begins, all pheromone concentrations are initialized to the same positive number, for example, let... This ensures that the initial search is not biased towards any particular pump operating speed.

[0056] S33b, according to the Only one ant has been identified as the previous one. Based on the pump operating speed, the current predicted reservoir water volume, the minimum continuous operating time of the pump, the minimum continuous shutdown time of the pump, and the maximum power change, the first... The set of feasible gears for each scheduling period .

[0057] In practical implementation, this applies to each water pump operating speed. First according to Calculate the candidate water extraction volumes, and then deduce and select the first one. The water volume in the reservoir after the first pump is set to its operating speed, if the first... The operating speed of a water pump can directly cause the water level in the reservoir to exceed the maximum allowable water level, or cause the pump power to change excessively during adjacent scheduling periods. Then the first Each water pump operating speed is added to the set of feasible speeds.

[0058] When the water pump is currently in continuous operation and the continuous operation time has not yet reached the specified time. The stop mode cannot be selected; when the water pump is currently in a continuous stop state and the continuous stop time has not yet reached the specified time. At this time, non-stop gears are not allowed to be selected.

[0059] When all pump operating speeds are excluded due to the water volume boundary of the reservoir, the water volume of the reservoir corresponding to each pump operating speed is estimated, and the water volume of the reservoir is calculated based on the predicted water volume. The intervals are sorted from smallest to largest distance, and the top 50% of pump operating speeds with the smallest distances are retained to form a temporary candidate speed set. The temporary candidate speed set still needs to meet the constraints of minimum continuous pump operation time, minimum continuous pump shutdown time, and maximum power change. Any water volume exceeding the limit generated by these temporary candidate speed sets is corrected in S304.

[0060] S33c, combining pheromone concentration, dynamic visibility, and feasible gear set, calculates the... In the nth iteration Only ants in the first Select the first scheduling period The formula for the transition probability of each pump's operating speed is as follows: In the formula, Indicates the first In the nth iteration Only ants in the first Select the first scheduling period The probability of switching between different operating speeds of a water pump; The pheromone influence index is used to control the strength of the influence of historical search experience on the transition probability. In this embodiment, it is preferably 1. The dynamic visibility impact index is used to control the intensity of the impact of the current scheduling conditions on the transfer probability. In this embodiment, it is preferably 2. This indicates the index for traversing the pump's operating speed. This function indicates the feasibility of gear selection. It takes the value 1 when the condition in parentheses is true, and 0 otherwise. Indicates the first In the nth iteration The scheduling period and the first The pheromone concentration corresponding to each pump operating speed Indicates the first In the nth iteration Only ants in the first Select the first scheduling period Dynamic visibility at each pump operating speed; Indicates the first In the nth iteration Only ants in the first Select the first scheduling period When the water pump is in its operating gear, the first Dynamic visibility at each pump operating speed; Indicates the first A set of feasible gears for each scheduling period.

[0061] S33d, perform roulette wheel sampling on the transition probabilities corresponding to each feasible pump operating position to determine the first... In the nth iteration Only ants in the first Pump operating speed during each scheduling period and set the water pump to the correct operating speed. Add to the current candidate scheduling path. After determining the pump operating speed, update the temporary reservoir water volume according to the water volume status deduction method in S301, and update the pump operating status, continuous running time, or continuous shutdown time according to the selected pump operating speed. Then proceed to the next step. For each scheduling period, recalculate the dynamic scheduling situation scalar, dynamic visibility, feasible gear set, and transition probability. Iterate through the first to the second... After the scheduling period, the first... Complete candidate scheduling path for each ant . No. In the next iteration Only ants together generate There are 10 candidate scheduling paths, among which 100 are available. This indicates the number of ants, which can be set based on the total number during the scheduling period and the number of selectable pump speeds.

[0062] S34, after the candidate scheduling path is generated, multi-constraint self-healing repair is performed. The critical time period causing constraint violations is located along the time direction, and the alternative with the lowest correction cost is selected from the candidate repair levels, transforming the infeasible scheduling path into a feasible candidate scheduling path. Each ant completes its task. After selecting the gear for each time period, the complete path is obtained, which is the candidate scheduling path. Only ants in the first Generate in the next iteration Several candidate paths are selected. The specific steps are as follows: S34a performs forward simulation on the complete candidate scheduling path. Starting from the first scheduling period, it calculates the pump power, water lifting volume, actual available irrigation water volume, water supply deficit, reservoir water volume, pump start / stop status, continuous running time and continuous downtime for each period, and records all constraint violation periods.

[0063] The constraint violation includes the water volume in the reservoir exceeding [a certain limit]. The water level in the reservoir is lower than The water supply deficit in a single scheduling period exceeds the allowable deficit threshold. Continuous running time is less than Continuous downtime is less than And the change in pump power during adjacent scheduling periods exceeds Allowable shortfall threshold You can take 20% of the predicted irrigation water demand for the corresponding scheduling period.

[0064] The earliest scheduling period in the forward simulation process where constraint violation occurs is denoted as... ,exist Candidate repair time periods were identified within the previous backtracking window, and the backtracking window length was set. (Set according to specific needs, preferably 8~24), the candidate repair time period range is: in, This indicates the index of candidate repair time periods.

[0065] Specifically: when the water level in the reservoir is lower than... Or water supply deficit exceeds In this case, priority should be given to selecting candidate repair periods within the retrospective window that allow for an increase in the water pump operating level. Candidate repair periods should have an increase in the water pump operating level and should prioritize meeting the requirement that the predicted available photovoltaic power is not less than [a certain value]. The dynamic dispatch status scalar is greater than 0.5 and the time-of-use electricity price is not higher than the benchmark electricity price. .in, The photovoltaic power screening ratio is preferably 0.3 to 0.6. After increasing the pump's operating speed, the remaining capacity of the reservoir corresponding to the candidate repair period should be sufficient to accommodate the increased water volume, to avoid premature water replenishment causing the reservoir's water level to exceed the limit later. ; When the water volume in the reservoir exceeds In this case, priority should be given to selecting candidate repair periods within the backtracking window that can reduce the operating level of the water pump, and priority should also be given to selecting periods with time-of-use electricity prices not lower than [a certain threshold]. The predicted available photovoltaic power is lower than Alternatively, scheduling periods with a dynamic scheduling status scalar not exceeding 0.5 can be used to reduce the impact on periods of high photovoltaic absorption and periods of urgent water lifting.

[0066] S34b, for each candidate repair period From the set of selectable speeds of the water pump The pump operating speeds that meet the constraints of pump power boundary, minimum continuous operating time, minimum continuous downtime, and maximum power change are selected to form a candidate repair speed set. Specifically: Replace the candidate pump's operating speed with the first one in the candidate scheduling path. The first scheduling period, and check the replacement period. The scheduling period up to the first The power variation between scheduling periods, and the duration of the continuous operating or continuous shutdown segment where the replacement operation occurs. The corresponding pump operating speed is added to the candidate repair speed set only if the replacement operation will not cause a new violation of the minimum start / stop time or the maximum power variation. .

[0067] For each pump operating position in the candidate repair position set, from the candidate repair time period The forward simulation was re-executed, and the results were obtained after adopting the candidate pump's operating speed during the original constraint violation period. Corresponding predicted water volume of the reservoir .in, Indicates the first The water pump operating speed for the first scheduling period is replaced by the first... After the water pump operates at the specified speed, at the... The predicted water volume of the reservoir is obtained at the end of each scheduling period.

[0068] S34c, based on the power change of the candidate pump operating speed relative to the original pump operating speed, and the degree of water flow exceeding the limit during the original constraint violation period after the re-forward simulation, the correction cost is calculated as follows: in, Indicates the first The scheduling period adopts the first... The cost of correction when the water pump is operating at a certain speed. Indicates the number before repair The operating speed of the raw water pump during each scheduling period. This represents the penalty coefficient when the water volume in the reservoir exceeds the maximum allowable water volume. This represents the penalty coefficient when the water level in the reservoir falls below the minimum allowable level. and The preferred value is 10; Denotes the positive part function, Indicates the first The water pump power corresponding to each water pump operating speed; Indicates the number before repair Raw water pump operating speed during each scheduling period The corresponding water pump power.

[0069] The first term of the correction cost is used to limit the degree of disturbance that the repair operation causes to the original candidate scheduling path, and the second term is used to penalize the reservoir water volume exceeding the limit. The third item in the section is used to punish those whose water level in the reservoir is lower than the required level. The latter two items are both 0 when the candidate repair operation can bring the predicted reservoir water volume within the allowable range.

[0070] Iterate through all candidate repair time periods and their corresponding repair speeds, and select the candidate repair time period and pump operating speed with the lowest correction cost, i.e.: in, This indicates the final selected repair period. This indicates the final selected repair setting. This represents the variable selection operation that minimizes the correction cost.

[0071] When multiple candidate repair operations have the same correction cost, the pump operation mode with higher dynamic visibility should be selected first; when the dynamic visibility is still the same, the time period furthest from the original constraint violation should be selected first. The nearest candidate repair period is selected to reduce the number of subsequent scheduling periods affected by the repair operation.

[0072] S34d, the candidate scheduling path is in the... The selected repair period corresponds to the original water pump operating speed, which is then replaced with the repair speed. And from the first The forward simulation will be re-executed at the start of each scheduling period. If constraint violations still occur subsequently, the earliest constraint violation period will be re-determined, and backtracking, candidate repair level selection, and minimum correction cost selection will continue. The maximum number of repairs is [number missing]. When all hard constraints are satisfied, the repaired candidate scheduling path is determined as a feasible candidate scheduling path; when hard constraints are still violated after the maximum number of repairs is reached, the corresponding candidate scheduling path is marked as an infeasible candidate scheduling path and does not participate in subsequent non-dominated solution set and pheromone update.

[0073] S35, perform multi-objective evaluation and Pareto ranking on all candidate scheduling paths, only use non-dominated solutions for pheromone enhancement, and improve the pheromone contribution of solutions in sparse regions by using congestion distance. The specific steps are as follows: S35a, calculate the... In this iteration, all ants generate multi-objective evaluation vectors for candidate scheduling paths, which are then self-healed by multiple constraints, and Pareto non-dominated sorting is performed. If a candidate scheduling path is not inferior to another candidate scheduling path on all objectives, and is superior on at least one objective, then that candidate scheduling path dominates the other candidate scheduling path. There is no non-dominated solution set consisting of candidate scheduling paths that are not dominated by other candidate scheduling paths. ,in, Indicates the first The set of non-dominated solutions obtained in the next iteration is used to retain candidate scheduling paths that are not comprehensively superior in multi-objective evaluation. The non-dominated solutions represent the optimal trade-offs among the four objectives in the current iteration, guiding the search toward convergence to the Pareto front.

[0074] S35b calculates the normalized congestion distance for candidate scheduling paths in the non-dominated solution set. (The remaining text appears to be incomplete and requires further context.) The normalized crowding distance in the target space is denoted as ,in, Represents candidate scheduling paths in the non-dominated solution set; This value is used to characterize the sparsity of the solution distribution around the candidate scheduling path. The larger the value, the sparser the target space region where the candidate scheduling path is located.

[0075] S35c performs minimum-maximum normalization on the target values ​​of operating cost, water supply deficit, curtailment loss, and shift switching, respectively, to obtain four normalized target values ​​for the candidate scheduling path. ,in, These correspond to the targets for operating costs, water supply deficit, solar power curtailment losses, and power grading switching, respectively.

[0076] Candidate scheduling paths are calculated based on the weighted sum of four normalized objective values. Corresponding basic pheromone increment : in, Indicates candidate scheduling paths The corresponding basic pheromone increment, Indicates the first The pheromone evaluation weights corresponding to each target This represents the stability constant for calculating pheromone increments, used to avoid a denominator of 0 when all four normalized target values ​​are 0. It can be taken as... When no target preference is specified, you can take... .

[0077] The smaller the weighted sum of the normalized objective values ​​of the candidate scheduling paths, the larger the increment of the basic pheromone, indicating that the candidate scheduling path has higher overall quality.

[0078] S35d involves the evaporation and enhancement of pheromones. The pheromone renewal method is represented as follows: in, Indicates the first In the nth iteration Time period and the The pheromone concentration corresponding to each pump's operating speed is used to guide the ants in constructing the scheduling path for the next round. Indicates the first The pheromone evaporation rate of the next iteration is used to control the decay rate of old path information, and its value ranges from 0.1 to 0.6. This represents the crowding distance enhancement coefficient, used to control the degree of enhancement of pheromone updates by the solution in sparse regions, and can be set to 0.5; This represents the path locator indicator function, when the candidate scheduling path is at the 0th position. Time period selection The value is 1 when the water pump is in its operating gear, and 0 otherwise. Indicates the candidate scheduling path at the th The pump operating speed corresponding to the time period.

[0079] S36, when the ant colony search reaches the maximum number of iterations. (Preferably 100 to 300), or when the Pareto front change is below the front change threshold for multiple consecutive rounds, output a set of non-dominated candidate scheduling schemes. Each candidate scheduling scheme includes the pump operation speed sequence, corresponding power allocation, reservoir water volume trajectory, and multi-objective evaluation results for the entire scheduling cycle. Finally, an execution scheme is selected from the final candidate scheduling scheme set based on actual operational preferences. When irrigation security is prioritized, the candidate scheduling scheme with the lowest water supply deficit target value is selected; when economic operation is prioritized, the candidate scheduling scheme with the lowest operating cost target value is selected; when renewable energy consumption is prioritized, the candidate scheduling scheme with the lowest curtailment loss target value is selected; when stable equipment operation is prioritized, the candidate scheduling scheme with the lowest speed switching target value is selected. If multiple objectives need to be balanced comprehensively, the objectives in the final candidate scheduling scheme set can be normalized and decision weights set, and the candidate scheduling scheme with the best comprehensive score can be selected. The reservoir water volume trajectory and power allocation diagram under the scheduling path of this invention are shown below. Figure 5 As shown, (a) is a graph showing the change in water volume and the safe range, and (b) is a stacked display of the available photovoltaic power, the power consumed by the water pump, the power purchased from the grid, and the power of the curtailed photovoltaic power.

[0080] The threshold for front change can be set based on changes in the number of non-dominated solutions, changes in the target space distance, or changes in the optimal target value. For example, the search can be stopped when the average change of the Pareto front is less than 1% of the threshold for front change in 10 consecutive iterations.

[0081] In this invention, the final candidate scheduling scheme set is not a single fixed scheduling result, but rather includes feasible schemes with multiple operational preferences. Through the synergistic effects of photovoltaic-hydraulic modal coupling feature mining, scheduling situation map construction, dynamic visibility guidance, Pareto pheromone updating, and multi-constraint self-healing repair, this invention can achieve closed-loop scheduling from historical time-series feature mining to full-cycle pump operation level optimization, thereby improving photovoltaic absorption rate, reducing electricity purchase costs, ensuring irrigation water supply, and reducing frequent pump start-ups and shutdowns.

[0082] This invention is applied to an online photovoltaic water pumping scheduling system. After completing the historical photovoltaic-water coupling feature mining and the water pump operation speed optimization model setting, the system updates the predicted available photovoltaic power sequence, the predicted irrigation water demand sequence, the real-time water volume of the reservoir, and the time-of-use electricity price sequence according to a fixed rolling cycle (e.g., 15 minutes). It then regenerates a candidate scheme set for the current scheduling cycle through the schemes in steps S1-3, selects the execution scheme from it, and issues only the speed command within the current step size. After execution, the actual feedback data is used as the input for the next round of rolling scheduling, achieving closed-loop correction. When the actual water volume is lower than the safety warning level, there is a sudden drop in photovoltaic power, or a sudden increase in water consumption, priority is given to ensuring water supply safety. Then, based on the electricity price and photovoltaic situation, a decision is made on whether to purchase electricity to supplement water and whether to delay non-emergency irrigation.

[0083] In this invention, online rolling scheduling execution enables the invention to continuously improve upon historical solar-water coupling patterns, current real-time status, and short-term future forecasts, rather than relying on a one-time static plan. Based on this, the feasibility and economy of the solar water pumping scheduling scheme can be maintained even under conditions of random fluctuations in solar power output, temporary changes in irrigation demand, and dynamic changes in reservoir water volume.

[0084] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0085] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of protection of the present invention.

Claims

1. A method for intelligent scheduling and power optimization of photovoltaic water pumping, characterized in that the steps include: include: Step S1: Collect historical data of the target agricultural irrigation area and preprocess the data; Step S2: Construct a modal coupling feature mining and scheduling situation map of photovoltaic power output and water demand time series; Step S3 involves improving the ant colony optimization algorithm by guiding the scheduling situation and using multi-constraint self-healing. The scheduling situation map is embedded into the dynamic visibility calculation, gear transfer probability, pheromone update and multi-constraint self-healing repair links of ant colony optimization, forming a multi-objective scheduling solution method suitable for photovoltaic water lifting tasks.

2. The method for intelligent scheduling and power optimization of photovoltaic water lifting according to claim 1, characterized in that, The data collected in S1 includes historical measured available photovoltaic power output, historical irrigation water demand, and the available photovoltaic power, irrigation water demand, real-time water volume in the reservoir, and time-of-use electricity price pump performance parameters predicted for the current scheduling cycle. The above data are used to construct a historical available photovoltaic power output sequence in chronological order. Historical irrigation water demand sequence and the available photovoltaic power sequence predicted for the current scheduling cycle. Irrigation water demand sequence Time-of-use electricity pricing series ; The specific steps of the preprocessing are as follows: S11, unify all multi-data samples into the same scheduling period; S12, correct abnormal data in the historical photovoltaic available output sequence and historical irrigation water demand sequence after unified scheduling time axis, including: 1) data missing correction; 2) abnormal mutation data correction.

3. The method for intelligent scheduling and power optimization of photovoltaic water lifting according to claim 1, characterized in that, The specific steps of S2 are as follows: S21, Adaptive variational mode decomposition is performed on the preprocessed historical photovoltaic available power output sequence and historical irrigation water demand sequence to obtain photovoltaic intrinsic mode components and water demand intrinsic mode components, and the center frequency of each photovoltaic intrinsic mode component and each water demand intrinsic mode component is recorded. S22, pair the photovoltaic intrinsic mode components with the water demand intrinsic mode components, and comprehensively measure the information sharing strength and center frequency proximity between each pair of modes to calculate the mutual information coupling coefficient between the photovoltaic intrinsic mode components and the water demand intrinsic mode components; S23, Selecting mutual information coupling coefficients greater than the selection threshold. The mode pairs are taken as the dominant resonant mode pairs. The dominant resonant mode pairs are reconstructed by weighting according to the coupling strength to obtain the dominant photovoltaic resonant sequence. Dominant water demand resonance sequence Harmony and light-water resonance components; S24, dynamically time-bending aligned the dominant photovoltaic resonance sequence and the dominant water demand resonance sequence, in order to Using the reference sequence, find the minimum matching path between the two sequences and establish an alignment mapping function. Divide the time index offset of each scheduling period by the number of scheduling periods per day to obtain the optical-water matching offset. S25, the predicted available photovoltaic power, predicted irrigation water demand, photovoltaic-water matching offset, reservoir water status and time-of-use electricity price in the current scheduling cycle are normalized to [0,1] by minimum-maximum, and the scheduling status scalar of each scheduling period is obtained by weighted nonlinear fusion, and then arranged in chronological order to form a scheduling status diagram.

4. The intelligent scheduling and power optimization method for photovoltaic water lifting according to claim 3, characterized in that, The specific steps of S22 are as follows: S22a, respectively, the first The photovoltaic intrinsic mode components, the first The modal values ​​of the intrinsic mode components of water demand over all scheduling periods constitute the photovoltaic intrinsic mode vector. Water demand intrinsic mode vector The photovoltaic intrinsic mode vector and the water demand intrinsic mode vector are adaptively binned, and the mutual information coupling coefficient is calculated based on the joint distribution after binning, as shown in the following formula: In the formula, Indicates the first The photovoltaic intrinsic mode component and the first The mutual information coupling coefficient between the intrinsic modal components of water demand. The coupling coefficient is used to calculate the stability constant. This represents the frequency bandwidth tolerance parameter. Represents the natural exponential function; This represents the absolute value operator. This represents the mutual information calculation function. Indicates the first Information entropy of photovoltaic intrinsic mode vectors Indicates the first Information entropy of the intrinsic mode vector of water demand Indicates the first The center frequency of each photovoltaic intrinsic mode component Indicates the first The center frequency of the intrinsic modal components of water demand; S22b, Calculate the frequency domain penalty weight based on the center frequency difference between the photovoltaic intrinsic mode components and the water demand intrinsic mode components. ; S22c, iterate through all photovoltaic intrinsic mode components and all water demand intrinsic mode components, calculate all mutual information coupling coefficients, and combine them to obtain the mutual information coupling coefficient matrix. .

5. The intelligent scheduling and power optimization method for photovoltaic water lifting according to claim 3, characterized in that, The specific steps of S23 are as follows: S23a, Screening for mutual information coupling coefficients greater than the screening threshold The mode pairs are considered as dominant resonance mode pairs, forming the dominant resonance mode pair set. and to The photovoltaic intrinsic mode components and water demand intrinsic mode components in the data are standardized with zero mean and unit variance, respectively, to obtain... Photovoltaic normalized intrinsic mode components at time and water demand normalized intrinsic mode components ; S23b, calculate the normalized coupling weights based on the mutual information coupling coefficients of each dominant resonance mode pair, and reconstruct the dominant photovoltaic resonance sequences according to the normalized coupling weights. Resonance sequence with dominant water demand ; S23c, by weighted fusion of the normalized intrinsic mode components of photovoltaics and the normalized intrinsic mode components of water demand, yields the photovoltaic-water resonance component. .

6. The intelligent scheduling and power optimization method for photovoltaic water lifting according to claim 1, characterized in that, The specific steps of S3 are as follows: S31 uses the pump operating speed sequence throughout the entire scheduling cycle as the ant's search path, searches for the Pareto approximate optimal solution set of the scheduling scheme, and constructs the scheduling path for each ant. ,in, Indicates the first The scheduling path constructed by an ant Indicates the first Only ants in the first The pump operating speed selected for the specified time period; S32, calculate the dynamic visibility of the scheduling situation guidance, for the first... In the nth iteration Only ants in the first The first scheduling period For each pump operating speed, the dynamic scheduling status scalar is combined with the ratio of pump power to rated power to calculate the status adaptation term; the electricity price sensitivity index is used as the power of the ratio of benchmark electricity price and time-of-use electricity price to calculate the electricity price adjustment term for each scheduling period; the absolute value of the difference between the current speed power and the selected speed power in the previous scheduling period and the negative value of the ratio of the power change tolerance parameter are used as the exponent of the natural constant e to calculate the speed switching smoothing term; finally, the dynamic visibility is calculated by combining the status adaptation term, power matching term, electricity price suppression term, and speed smoothing term. S33, calculate the transfer probability of each pump's operating gear based on the current pheromone concentration, dynamic visibility, and gear feasibility constraints, and generate a complete scheduling path for each time period according to the probability. S34. After the candidate scheduling path is generated, perform multi-constraint self-healing repair, locate the critical period that causes the constraint violation along the time direction, and select the alternative position with the lowest correction cost from the candidate repair positions to transform the infeasible scheduling path into a feasible candidate scheduling path. S35 performs multi-objective evaluation and Pareto ranking on all candidate scheduling paths, only uses non-dominated solutions for pheromone enhancement, and improves the pheromone contribution of sparse region solutions by crowding distance. S36, when the ant colony search reaches the maximum number of iterations. If the Pareto frontier changes are below the frontier change threshold for multiple consecutive rounds, output a set of non-dominated candidate scheduling schemes. Each candidate scheduling scheme includes the pump operation speed sequence, corresponding power allocation, water volume trajectory of the reservoir, and multi-objective evaluation results for the entire scheduling cycle. Finally, the execution scheme is selected from the final candidate scheduling scheme set based on actual operation preferences.

7. The intelligent scheduling and power optimization method for photovoltaic water lifting according to claim 6, characterized in that, The specific steps of S31 are as follows: S31a, read the... Pump power and predicted reservoir water volume for time-scheduled routes and the Predicted water volume in reservoirs for different time periods , will the Water flow rate corresponding to the pump operating speed during the time period and scheduling interval Multiply by the product to obtain the water extraction volume; then use the predicted irrigation water demand. The outflow of water is deducted, and the constraint violation of the scheduling path is calculated through state deduction. All constraint violations are recorded. Specifically: if the predicted water volume of the reservoir is lower than the minimum allowable water volume of the reservoir, a water supply shortage or a violation of the lower limit of the water level will occur. If the predicted water volume of the reservoir is higher than the maximum allowable water volume of the reservoir, an overflow risk or a violation of the upper limit of the water level will occur. S31b, the target values ​​of operating cost, water supply deficit, curtailment loss and gear switching are concatenated into a four-dimensional vector to obtain the multi-objective evaluation vector of each scheduling path; S31c specifies the design capacity of the water storage tank and the minimum continuous operating time of the water pump. Minimum continuous downtime of water pump and maximum gear change The constraint is used as a hard constraint to verify each path on a time-by-time basis, and to filter out scheduling paths that are not feasible to actually execute.

8. The intelligent scheduling and power optimization method for photovoltaic water lifting according to claim 6, characterized in that, The specific steps of S33 are as follows: S33a, Establish the pheromone matrix corresponding to the scheduling period and the pump operating speed; S33b, according to the Only one ant has been identified as the previous one. Based on the pump operating speed, the current predicted reservoir water volume, the minimum continuous operating time of the pump, the minimum continuous shutdown time of the pump, and the maximum power change, the first... The set of feasible gears for each scheduling period ; S33c, combining pheromone concentration, dynamic visibility, and feasible gear set, calculates the... In the nth iteration Only ants in the first Select the first scheduling period The formula for the transition probability of each pump's operating speed is as follows: In the formula, Indicates the first In the nth iteration Only ants in the first Select the first scheduling period The probability of switching between different operating speeds of a water pump; This indicates the pheromone influence index. The dynamic visibility impact index indicates the impact of visibility. This indicates the index for traversing the pump's operating speed. This indicates the feasibility indicator function for gear selection. Indicates the first In the nth iteration The scheduling period and the first The pheromone concentration corresponding to each pump operating speed Indicates the first In the nth iteration Only ants in the first Select the first scheduling period Dynamic visibility at each pump operating speed; Indicates the first In the nth iteration Only ants in the first Select the first scheduling period When the water pump is in its operating gear, the first Dynamic visibility at each pump operating speed; Indicates the first A set of feasible gears for each scheduling period; S33d, perform roulette wheel sampling on the transition probabilities corresponding to each feasible pump operating position to determine the first... In the nth iteration Only ants in the first Pump operating speed during each scheduling period and set the water pump to the correct operating speed. Add to the current candidate scheduling path; after determining the pump operating speed, update the temporary reservoir water volume according to the water volume status extrapolation method, and update the pump operating status, continuous running time or continuous shutdown time according to the selected pump operating speed, and then proceed to the next step. For each scheduling period, recalculate the dynamic scheduling situation scalar, dynamic visibility, feasible gear set, and transition probability, traversing from the 1st to the th... After the scheduling period, the first... Complete candidate scheduling path for each ant .

9. The intelligent scheduling and power optimization method for photovoltaic water lifting according to claim 6, characterized in that, The specific steps of S34 are as follows: S34a performs forward simulation on the complete candidate scheduling path. Starting from the first scheduling period, it calculates the pump power, water lifting volume, actual available irrigation water volume, water supply deficit, water storage tank volume, pump start-stop status, continuous running time and continuous downtime for each period, and records all constraint violation periods. The earliest scheduling period in the forward simulation process where constraint violation occurs is denoted as... ,exist Candidate repair time periods were identified within the previous backtracking window, and the backtracking window length was set. ; S34b, for each candidate repair period From the set of selectable speeds of the water pump The pump operating speeds that meet the constraints of pump power boundary, minimum continuous operating time, minimum continuous downtime, and maximum power change are selected to form a candidate repair speed set. ; S34c: Based on the power change of the candidate pump operating position relative to the original pump operating position, and the degree of water volume exceeding the limit during the original constraint violation period after the re-forward simulation, the correction cost is calculated through the penalty coefficient. All candidate repair periods and their candidate repair positions are traversed, and the candidate repair period and pump operating position with the minimum correction cost are selected. S34d, the candidate scheduling path is in the... The selected repair period corresponds to the original water pump operating speed, which is then replaced with the repair speed. And from the first The forward simulation will be re-executed at the start of each scheduling period. If there are still constraint violations, the earliest constraint violation period will be re-determined and the backtracking location, candidate repair level selection, and minimum correction cost selection will continue.

10. The intelligent scheduling and power optimization method for photovoltaic water lifting according to claim 6, characterized in that, The specific steps of S35 are as follows: S35a, calculate the... In each iteration, all ants generate multi-objective evaluation vectors for candidate scheduling paths after multi-constraint self-healing repair. Pareto non-dominated sorting is then performed to determine if any candidate scheduling path dominates another candidate scheduling path. Candidate scheduling paths that are not dominated by other candidate scheduling paths form a non-dominated solution set. ; S35b, calculate the normalized congestion distance for candidate scheduling paths in the non-dominated solution set, and then... The normalized crowding distance in the target space is denoted as ; S35c performs minimum-maximum normalization on the target values ​​of operating cost, water supply deficit, curtailment loss, and shift switching, respectively, to obtain four normalized target values ​​for candidate scheduling paths; then, the candidate scheduling path is calculated based on the weighted sum of the four normalized target values. Corresponding basic pheromone increment ; S35d involves the evaporation and enhancement of pheromones. The pheromone renewal method is represented as follows: in, Indicates the first In the nth iteration Time period and the The pheromone concentration corresponding to each pump operating speed Indicates the first Pheromones evaporation rate in each iteration. Indicates the crowding distance amplification coefficient. This indicates the path / gear indicator function. Indicates the candidate scheduling path at the th The pump operating speed corresponding to the time period.

Images