CN104699979A - Urban lake and reservoir algal bloom chaos time sequence predication method based on complicated network - Google Patents

Abstract

The invention discloses an urban lake and reservoir algal bloom chaos time sequence predication method based on a complicated network and belongs to the technical field of environmental engineering. A chaos property is checked in a lake and reservoir bloom generation process, and a bloom predication method based on a chaos time sequence is given; aiming at solving the problems that almost existing bloom predication can be used for predicating single factors and the predication precision is not high and the like and aiming at a bloom generation process characteristic factor time sequence based on chaos properties, statistical characteristic G parameters in the bloom generation process by adopting a complicated network method, and a G parameter time sequence is predicated by the chaos time sequence predication method, so that the multi-factor chaos time sequence predication process predication is realized and the predication precision is improved; effective reference evidences are provided for an environment-friendly department, and important treatment and prevention effects on the protection and improvement of lake and reservoir environments are realized.

Description

Based on the urban lake storehouse algal bloom Study on prediction technology of chaotic series of complex network

Technical field

The present invention relates to a kind of wawter bloom Forecasting Methodology, belong to field of environment engineering technology, specifically, refer on the basis of further investigation wawter bloom genesis mechanism, chaotic characteristic inspection is carried out to it, then in conjunction with the statistical nature of complex network, build urban lake storehouse algal bloom Chaotic time series forecasting model to explore the effective way improving wawter bloom precision of prediction.

Background technology

Along with the fast development of economic society, water environmental problems becomes increasingly conspicuous, and wherein body eutrophication is one of water environmental problems that is the most general, that have the greatest impact.Because China Hu Ku receives the holophytic nutrition salt such as excessive nitrogen, phosphorus in recent years, change the trophic structure of water body, make algae and other hydrophyte distorted proliferation, cause water quality deterioration, water transparency and dissolved oxygen DO decline, the body eutrophication problem of fish and other biological mass mortality, and then cause algal bloom to occur.Body eutrophication can destroy aquatic ecosystem, causes wawter bloom, and release Algae toxins, causes higher organism perish, the destroyed area ecologic equilibrium, restriction Regional Economic Development.

Wawter bloom is the typical performance of body eutrophication, is the coefficient results such as nutritive salt factor, environmental factor, biodyne and kinetic factor, and between each key element, relation is complicated.

Due to wawter bloom occur complicated mechanism, influence factor is more, thus to its predict be all the time wawter bloom administer and preventing and controlling in a difficult point.In the last few years, along with going deep into of research, many models set up based on intelligent method are applied in the middle of wawter bloom prediction, as regression model, neural network model etc.

Current in the multiple Forecasting Methodology of wawter bloom, be all that the single sign factor by describing wawter bloom phenomenon to chlorophyll etc. is predicted mostly.It is incomplete for describing wawter bloom phenomenon by single factors, and therefore the prediction of single factors is also incomplete for wawter bloom prediction.

In addition, because wawter bloom generative process is a complication system, therefore need to adopt analysis of complex system method to be described.Complex network is that entity interactional in complication system is abstracted into node, is reflected the interaction of each entity in complication system by internodal interaction.Therefore, complex network can be described as the model in reality on visible model, and obtains visible result by mathematical operation, and the feature of more complete description complication system, therefore complex network can apply to the modeling of wawter bloom generative process.

In addition, a lot of complicated system often has nonrandom but seemingly random chaos characteristic, and therefore the system studying this high complexity that appears as of chaology provides new thinking.Chaotic time series forecasting is an important applied field and the study hotspot of chaology, and its key idea is exactly that time lag is reproduced, and by introducing time delay and Embedded dimensions, One-dimension Time Series is transformed into multidimensional phase space to rebuild motive power system.The phase space of rebuilding maintains original geometry, i.e. topological structure, and has same dynamics.In actual wawter bloom generative process, due to each characteristic factor interphase interaction, make it often have some seemingly random features, therefore need carry out chaotic characteristic inspection for this complication system of wawter bloom generative process, and carry out modeling and forecasting for the characteristic factor sequential with chaotic characteristic.

Summary of the invention

The present invention carries out chaotic characteristic inspection to storehouse, lake wawter bloom generative process, and the wawter bloom Forecasting Methodology provided based on chaos time sequence, object is the problems such as the large multipair single factors prediction of solution existing wawter bloom prediction and precision of prediction are not high, for the wawter bloom generative process characteristic factor sequential with chaotic characteristic, adopt the statistical nature G parameter of complex network method construct wawter bloom generative process, by the Forecasting Methodology of chaos time sequence, G parameter time series is predicted, thus realize the prediction of multifactorial wawter bloom generative process, improve precision of prediction, for environmental administration provides effective reference frame, important preventive and therapeutic effect is played to the protection of storehouse, lake water environment and improvement.

In the present invention, the characteristic factor relevant with wawter bloom phenomenon is divided into two kinds: a kind of is affect the characteristic factor that wawter bloom occurs, and such as total nitrogen, total phosphorus, pH value, dissolved oxygen DO, temperature, intensity of illumination etc., hereinafter referred to as influence factor; Another kind is the characteristic factor characterizing wawter bloom generation, such as chlorophyll concentration, algae density etc., hereinafter referred to as sign factor.

Provided by the invention based on complex network urban lake storehouse algal bloom Study on prediction technology of chaotic series, comprise following five steps:

Step one, urban lake storehouse algal bloom generative process chaotic characteristic are checked;

1, wawter bloom characterizes choosing of factor;

Wawter bloom characterizes because have chlorophyll concentration, algae density etc., wherein algae density is indirect inspection factor, for improving seasonal effect in time series confidence level and computational accuracy, the present invention adopts chlorophyll concentration to carry out chaotic characteristic inspection as wawter bloom sign factor to wawter bloom generative process.

2, chlorophyll time series phase space reconfiguration;

Wawter bloom generative process is by coefficient results of many factors such as nutritive salt factor, environmental factor, biodynes, show seem random but and nonrandom characteristic.For this complication system of wawter bloom generative process, the evolution of arbitrary component all by other with it interactional component determined.The object of phase space reconfiguration is exactly in higher-dimension phase space, recover to present the attractor of nonlinear motion rule in wawter bloom generative process, thus excavates information implicit in arbitrary component evolutionary process.Therefore, from time series, construct the vector of one group of m dimension, prop an embedded space, as long as the dimension of embedded space is abundant, just can recover the dynamics that this complication system of wawter bloom generative process is original, the phase space of reconstruct has similar geometrical property and information characteristic to original wawter bloom generative process power system.

To given time series x (t), t=1,2 ... N}, time delay is τ, and Embedded dimensions is m, if m < is N, the N number of data in time series opened up and prolongs into the vector that N-(m-1) τ m ties up phase space.

X(t)＝(x(t),x(t+τ),…x(t+(m-1)τ)),t＝1,…,M,M＝N-(m-1)τ

Wherein, X (t) is the phase space vector after reconstruct, and x (t) is the chlorophyll concentration value of not monitoring in the same time, and t is the seasonal effect in time series sampling time, and N is number of samples.

3, chlorophyll time series chaotic characteristic inspection;

Chlorophyll time series correctly can be checked whether to have chaotic characteristic, whether can go to analyze chlorophyll time series by chaology by directly determining.This patent adopts largest Lyapunov exponent method to judge whether algal bloom characteristic factor time series is chaos time sequence.As index Lyapunov<0, show the phase volume of this complex power system of wawter bloom generative process on corresponding dimension direction be shrink, stable.Otherwise, if Lyapunov>0 on certain direction, show that the phase volume of this complex power system of wawter bloom generative process constantly expands and folds on corresponding dimension direction, path contiguous in attractor becomes more and more uncorrelated, thus can not predict its long-term evolution behavior, illustrate that now wawter bloom generative process develops and present chaotic characteristic.That is, Lyapunov>0 can be used as strange attractor existence in wawter bloom generative process, shows that the wawter bloom generative process corresponding with time series has chaotic characteristic.

Step 2, generate the structure of model of cognition based on the algal bloom of complex network;

1, the directed networks model of wawter bloom forming process is built;

Because storehouse, lake water body is an open complication system, existing algal bloom formation mechanism modeling method cannot exactly, quantitatively breakout of water bloom is described during relation between water nutrition, between nutritive salt and external environment etc.The thought of complex network is that entity interactional in complication system is abstracted into node, is reflected the interaction of each entity in complication system by internodal interaction.Therefore, storehouse, the lake water body system of complexity is abstracted into a complex network by the present invention, and the major influence factors affecting algal bloom generation is abstracted into network node, forms network point set V (V={v ₁, v ₂..., v _n), v _ibe used for expression i-th network node, network node adds up to n; Relation between each major influence factors is abstracted into the limit of network, forms limit collection E (E={e ₁, e ₂..., e _m), e _ibe used for expression i-th limit, total limit number is m, and limits all in the collection E of limit all in pointed set V two nodes correspond.The network topological diagram that this point set V that is abstracted into by storehouse, lake water body system major influence factors and mutual relationship thereof and limit collection E forms constitutes directed networks MODEL C N=(V, E), for representing storehouse, lake algae and water wawter bloom Formation and characteristics.

2, calculation of complex network characterization parameter;

The data characteristics of complication system shows often by the statistical nature of the complex network model of its correspondence, and therefore, the statistical property of complex network becomes the focus of a lot of scholar.At present, the statistical property that can calculate mainly comprises average path length L, node betweenness B _i, cluster coefficients C, bee-line d _mindeng, wherein i represents node serial number.Complexity between the non-linear effects factor of breakout of water bloom to be interacted and effect degree can be able to more complete sign by the statistical property of complex network.3, wawter bloom generative process Statistic Characteristic of Complex Network G parameter is built;

The effect in a network of different characteristic parameter is different, in order to characterize the overall characteristic in wawter bloom generative process, builds the model of the crucial degree σ of node, and the statistical nature G parameter of calculation of complex network, G=f (σ).

Step 3, G parameter time series chaotic characteristic are checked;

To the time series { g (t) of statistical nature G parameter, t=1,2, N} carries out phase space reconfiguration, find optimal delay time τ and Embedded dimensions m, wherein m < N, then can open up the N number of data in time series and prolong into the vector that N-(m-1) τ m ties up phase space

G(t)＝(g(t),g(t+τ),…g(t+(m-1)τ)),t＝1,…,M,M＝N-(m-1)τ

Wherein, G (t) is the phase space vector after the reconstruct of G parameter time series, and g (t) is the G parameter time series value do not obtained in the same time.

Then calculate maximum Lyapunov exponent, and check the chaotic characteristic of G parameter time series.If G parameter time series is chaos, then the Local prediction of conventional chaos time sequence can be adopted to predict G parameter time series.

Step 4, to predict based on the G parameter time series of chaos time sequence;

Here, G parameter is the sign of the abstract network overall permanence that this complication system of wawter bloom generative process is corresponding, therefore, carry out predicting to G parameter time series and effectively can solve the prediction of existing wawter bloom only for the forecasting problem of single sign factor, realize the object of wawter bloom integrated forecasting, and improve precision of prediction.

Herein, its essence of the adding-weight one-rank local-region method of employing be using last vectorial G (M) in phase space reconstruction as prediction central point, utilize other vector and G (M) between distance, simulate first-order linear fitting coefficient.Then, first-order linear fit approach just can be utilized to approach future evolution phase point, obtain the predicted value of next step evolution phase point:

G(M _j+1)＝a+bG(M _j)

Wherein a, b are first-order linear fitting coefficient, M _j(j=1,2 ..., q) represent the sequence number with prediction starting point G (M) adjacent nearer jth neighbor point vector, q then for specify with prediction starting point G (M) local adjacent nearer sequence of vectors number.

The invention has the advantages that:

1, the present invention is checked by chaotic characteristic, confirms that storehouse, lake wawter bloom generative process has chaotic characteristic, proposes the method predicted wawter bloom sign factor sequential based on chaos time sequence.

2, the present invention is directed to the deficiency of wawter bloom generative process single factors Forecasting Methodology, theoretical in conjunction with complex network related statistical, propose and adopt the comprehensive statistics feature G parameter of multiple characteristic factor as new forecasting object, quantificational description is carried out to this abstract complex network of wawter bloom generative process, for the prediction of wawter bloom generative process multi-factor comprehensive provides possibility.

Accompanying drawing explanation

Fig. 1 is the process flow diagram of the urban lake storehouse algal bloom Study on prediction technology of chaotic series that the present invention is based on complex network;

Fig. 2 is that the complex network algal bloom affected by composite factor is formed to network model;

Fig. 3 is chlorophyll time series normalized set result;

Fig. 4 is chlorophyll time series largest Lyapunov exponent after reconstruct;

Fig. 5 is that Statistic Characteristic of Complex Network G parameter and chlorophyll contrast;

Fig. 6 is the normalized set result of G parameter time series;

Fig. 7 is the largest Lyapunov exponent of G parameter time series after reconstruct;

Embodiment

Below in conjunction with accompanying drawing, the present invention is described in further detail.

The urban lake storehouse algal bloom Study on prediction technology of chaotic series of complex network provided by the invention, flow process as shown in Figure 1, concrete steps are as follows:

Step one, urban lake storehouse algal bloom generative process chaotic characteristic are checked;

1, wawter bloom characterizes choosing of factor;

Owing to only having chlorophyll directly to measure in sign factor, therefore adopt chlorophyll concentration to characterize factor as wawter bloom and chaotic characteristic inspection is carried out to wawter bloom generative process.

2, chlorophyll time series phase space reconfiguration;

The present invention adopts conventional C-C algorithm to carry out phase space reconfiguration to chlorophyll time series:

According to C-C algorithm construction statistic, according to calculate obtain feature determination optimum delay time τ and the smallest embedding dimension number m of corresponding statistic, thus make the phase space of reconstruct farthest reflect the dynamic characteristic of former urban lake storehouse algal bloom generative process system.

3, chlorophyll time series chaotic characteristic inspection;

The present invention adopts small data sets arithmetic to calculate chlorophyll seasonal effect in time series maximum Lyapunov exponent after phase space reconstruction:

First, to the state vector of the M in phase space reconstruction ask for each neighbor point to t (t=1,2 ... M) distance after individual discrete time step, then maximum Lyapunov exponent can by each phase point and its nearest neighbor in orbit on average disperse rate estimates.Changed by natural logarithm, can obtain each neighbor point to spacing natural logarithm value along with discrete time step t change a series of straight lines, the slope of every bar straight line is roughly suitable with maximum Lyapunov exponent, be averaged by this slope of least square fitting, try to achieve maximum Lyapunov exponent.

Step 2, generate the structure of model of cognition based on the algal bloom of complex network;

1, the directed networks model of wawter bloom forming process is built;

According to urban lake storehouse wawter bloom formation mechenism feature, 7 key influence factors such as total phosphorus (TP), total nitrogen (TN), temperature (T), pH value, dissolved oxygen DO (DO), illumination (I), chlorophyll (chl_a) are formed point set V, mutual relationship between influence factor forms limit collection E, wherein nodes n=7, limit number m=14.The algal bloom building complex network is formed to network model as shown in Figure 2, and visible algal bloom generative process is subject to the combined influence of the every water body factor in urban lake storehouse and weather element, can abstractly be corresponding network model.

2, calculation of complex network characterization parameter;

Be formed to network model according to the algal bloom of complex network, routinely formulae discovery Statistic Characteristic of Complex Network parameter, need to obtain average path length L, node betweenness B _i, cluster coefficients C and bee-line d _minthese key parameters.

3, breakout of water bloom Statistic Characteristic of Complex Network G parameter model is built;

For better embodying the effect of each node in complex network, build the crucial degree model σ of node _i'for:

${\mathrm{\&sigma;}}_{i^{\&prime;}}=\frac{{\mathrm{\&alpha;}}^{+}{\mathrm{\&lambda;}}_i^{+}+{\mathrm{\&alpha;}}^{-}{\mathrm{\&lambda;}}_i^{-}}{d_{\min }}\&times;B_i$

In formula, d _mini.e. node v _ito the bee-line that chlorophyll a calculates, λ _ifor node degree, be divided into in-degree and out-degree b _ifor node betweenness, α ⁺, α ^-for weights, span is the real number of 0 to 1, and i represents node serial number, i=1,2 ..., 7.

Consider that average path length and cluster coefficients are on the impact of network, carry out second-order correction to the crucial degree model of node, correction model is:

${\mathrm{\&sigma;}}_i=(1-\mathrm{\&beta;}){\mathrm{\&sigma;}}_{i^{\&prime;}}+\mathrm{\&beta;}\frac{C_i\underset{j=1,j\&NotEqual;i}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&sigma;}}_{j^{\&prime;}}}{(n-1)L}$

In formula, C _ifor node v _icluster coefficients, represent node v _iwith the degree of coupling of other nodes in complex network, L is the average path length of network, j=1,2 ..., 7, β is weights, and span is the real number of 0 to 1.

Consider L, B _i, C _i, d _minand σ _ibe the key parameter that wawter bloom forms complex network characteristic, then build algal bloom and generate identification G parameter model:

$G=\exp (\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&sigma;}}_i^{\&prime;}\&times;d_{i\min })$

In formula, σ _i ^,representation node v _icrucial degree model σ _iwith node v _ithe product of design parameter value, i.e. σ _i ^,=c _i× σ _i, c _irepresent node v _idesign parameter value, n is nodes.Described concrete G parameter model comprises average path length L, node betweenness B _i, cluster coefficients C and bee-line d _minthese key parameters and crucial degree model σ _i.

Step 3, G parameter time series chaotic characteristic are checked;

C-C method in refer step one is carried out state space reconstruction to G parameter time series and is judged the chaos attribute of G parameter time series by calculating maximum Lyapunov exponent.

Step 4, to predict based on the G parameter time series of chaos time sequence;

Adopt adding-weight one-rank local-region method to predict statistical nature G parameter time series, mainly utilize the first-order linear fitting coefficient a in first-order linear fitting formula, b obtains the prediction of next step evolution phase point:

G(M _j+1)＝a+bG(M _j),j＝1,2,…,q

Then in predictive vector, last one dimension component is exactly G parameter time series sequence prediction value.

Embodiment 1:

Step one, urban lake storehouse algal bloom generative process chaotic characteristic are checked;

The data that model adopts are the breakout of water bloom Monitoring Data under glass sunlight house simulating natural condition in 2007, and water sample picks up from Yuyuan Lake Park, Beijing, and it is the important component part of Beijing's second ring's stream, has good representativeness.

7 wawter bloom characteristic factors in glass sunlight house year August in June, 2007 to 2007 are monitored, specifically in table 1.

Table 1 wawter bloom characteristic factor monitoring list

Title

PH value

Illumination

Temperature

Total nitrogen

Total phosphorus

Dissolved oxygen DO

Chlorophyll

Unit

Nothing

Lx

℃

mg/L

mg/L

mg/L

mg/L

Its Determination of Chlorophyll is that wawter bloom characterizes factor, and remaining 6 characteristic factor is wawter bloom influence factor.Monitoring equipment have recorded altogether 45 days every 1 hour wawter bloom characteristic factor data of 1066 groups, and carry out phase space reconfiguration to its Determination of Chlorophyll time series, the result of calculation of statistic as shown in Figure 3, can draw optimal delay time τ by Fig. 3 _d=3, embed window width τ _w=16, by τ _w=(m-1) τ draws m=6.

According to the phase space reconstructed, calculate largest Lyapunov exponent, result as shown in Figure 4.Result of calculation shows, and average gradient is greater than 0, i.e. Lyapunov>0, illustrates that chlorophyll time series has chaos attribute.

Step 2, wawter bloom generative process Statistic Characteristic of Complex Network G parameter model build;

Carry out the calculating of Complex Networks Feature parameter according to 109 groups of data of network topology to glass sunlight house in June, 2007, wherein, between the bee-line on directed complex networks limit and node, shortest path result of calculation is in shown in table 2,3.

The bee-line on table 2 breakout of water bloom identification directed complex networks limit

The internodal shortest path of table 3 breakout of water bloom identification directed complex networks

Analyze parameters in complex network in conjunction with real data, result of calculation is in table 4.

Table 4 wawter bloom generates each parameter value identifying complex network

Utilize the statistical nature G parameter model built, in conjunction with real data, calculate feature G parameter time series, and compared with chlorophyll time series, result as shown in Figure 5.As shown in Figure 5, the algal bloom statistical nature G parameter formed and the value characterizing the chlorophyll a that wawter bloom is formed has correlativity clearly.Show that statistical nature G parameter can reflect algal bloom forming process well, compare chlorophyll simultaneously and additionally provide abundanter information.The change of feature G parameter is ahead of chlorophyllous change, and from analysis, the numerical model that constructed wawter bloom forms the statistical nature G parameter of complex network can react wawter bloom formational situation preferably, can identify wawter bloom phenomenon and breakout of water bloom trend quickly.When the content of the key factor affecting breakout of water bloom in water body changes or changes trend, there is certain hysteresis quality in the change due to chlorophyll a, the situation of breakout of water bloom can not be reflected rapidly, but by the calculating to G parameter model, situation of change and the variation tendency of water body Determination of Chlorophyll a can be identified fast, thus take prevention and cure measures to provide reliable guarantee in time for relevant department.

Step 3, G parameter time series chaotic characteristic are checked;

To in June, 2007-July, the statistical nature G parameter time series of 605 groups of data carries out phase space reconfiguration, and the statistic of calculating and largest Lyapunov exponent are as shown in Figure 6, Figure 7.

Lyapunov exponent is greater than 0, illustrates that Statistic Characteristic of Complex Network G parameter time series has chaos attribute.

Step 4, to predict based on the G parameter time series of chaos time sequence;

Adopt adding-weight one-rank local-region method to predict statistical nature G parameter, predicting the outcome compares with actual value, and result is as shown in table 5.

Table 5 statistical nature parameter G predicted value and measured value contrast

Predict the outcome and there is slight error with measured value.Illustrate that the time forecasting methods based on chaology have good prediction effect in breakout of water bloom prediction thus, by the time series forecasting to Complex Networks Feature G parameter, change the history of breakout of water bloom unitary variant prediction, achieve system prediction.

Claims (4)

1., based on the urban lake storehouse algal bloom Study on prediction technology of chaotic series of complex network, it is characterized in that: comprise the following steps,

Step one, urban lake storehouse algal bloom generative process chaotic characteristic are checked;

Adopt chlorophyll concentration to characterize factor as wawter bloom and chaotic characteristic inspection is carried out to wawter bloom generative process;

To given time series x (t), t=1,2 ... N}, time delay is τ, and Embedded dimensions is m, if m < is N, the N number of data in time series opened up and prolongs into the vector that N-(m-1) τ m ties up phase space:

X(t)＝(x(t),x(t+τ),…x(t+(m-1)τ)),t＝1,…,M,M＝N-(m-1)τ

Wherein, X (t) is the phase space vector after reconstruct, and x (t) is the chlorophyll concentration value of not monitoring in the same time, and t is the seasonal effect in time series sampling time, and N is number of samples;

Largest Lyapunov exponent method is adopted to judge whether algal bloom characteristic factor time series is chaos time sequence, as index Lyapunov<0, show the phase volume of this complex power system of wawter bloom generative process on corresponding dimension direction be shrink, stable; Otherwise, if Lyapunov>0 on certain direction, show that wawter bloom generative process develops and present chaotic characteristic;

Step 2, generate the structure of model of cognition based on the algal bloom of complex network;

(1) the directed networks model of wawter bloom forming process is built;

Storehouse, the lake water body system of complexity is abstracted into a complex network, the major influence factors affecting algal bloom generation is abstracted into network node, form network point set V (V={v ₁, v ₂..., v _n), v _ibe used for expression i-th network node, network node adds up to n; Relation between each major influence factors is abstracted into the limit of network, forms limit collection E (E={e ₁, e ₂..., e _m), e _ibe used for expression i-th limit, total limit number is m, and limits all in the collection E of limit all in pointed set V two nodes correspond; The network topological diagram that this point set V that is abstracted into by storehouse, lake water body system major influence factors and mutual relationship thereof and limit collection E forms constitutes directed networks MODEL C N=(V, E), for representing storehouse, lake algae and water wawter bloom Formation and characteristics;

(2) calculation of complex network characterization parameter;

Described Complex Networks Feature parameter comprises average path length L, node betweenness B _i, cluster coefficients C and bee-line d _min, wherein i represents node serial number;

(3) wawter bloom generative process Statistic Characteristic of Complex Network G parameter is built;

Step 3, G parameter time series chaotic characteristic are checked;

To time series { g (t), t=1,2 of G parameter, N} carries out phase space reconfiguration, finds optimal delay time τ and Embedded dimensions m, wherein m < N, then the N number of data in time series can be opened up and prolong into the vector that N-(m-1) τ m ties up phase space

G(t)＝(g(t),g(t+τ),…g(t+(m-1)τ)),t＝1,…,M,M＝N-(m-1)τ

Wherein, G (t) is the phase space vector after the reconstruct of G parameter time series, and g (t) is the G parameter time series value do not obtained in the same time.

Then calculate maximum Lyapunov exponent, and check the chaotic characteristic of G parameter time series; If G parameter time series is chaos, then the Local prediction of conventional chaos time sequence is adopted to predict G parameter time series;

Step 4, to predict based on the G parameter time series of chaos time sequence;

The adding-weight one-rank local-region method adopted carries out the prediction of G parameter time series.

2. the urban lake storehouse algal bloom Study on prediction technology of chaotic series based on complex network according to claim 1, it is characterized in that: adding-weight one-rank local-region method in step 4, as prediction central point using last vectorial G (M) in phase space reconstruction, utilize the distance between other vector and G (M), simulate first-order linear fitting coefficient; Utilize first-order linear fit approach to approach future evolution phase point, obtain the predicted value of next step evolution phase point:

G(M _j+1)＝a+bG(M _j)

Wherein a, b are first-order linear fitting coefficient, M _jrepresent the sequence number with prediction starting point G (M) adjacent nearer jth neighbor point vector, j=1,2 ..., q, q be specify with prediction starting point G (M) local adjacent nearer sequence of vectors number.

3. the urban lake storehouse algal bloom Study on prediction technology of chaotic series based on complex network according to claim 1, it is characterized in that: described point set V comprises 7 key influence factors, be respectively total phosphorus TP, total nitrogen TN, temperature T, pH value, dissolved oxygen DO DO, illumination I, Chlorofucsin hl_a.

4. the urban lake storehouse algal bloom Study on prediction technology of chaotic series based on complex network according to claim 1, is characterized in that: build wawter bloom generative process Statistic Characteristic of Complex Network G parameter model, be specially:

First the crucial degree model σ of node is built _i'for:

${\mathrm{\&sigma;}}_{i^{\&prime;}}=\frac{{\mathrm{\&alpha;}}^{+}{\mathrm{\&lambda;}}_i^{+}+{\mathrm{\&alpha;}}^{-}{\mathrm{\&lambda;}}_i^{-}}{d_{\min }}\&times;B_i$

In formula, d _mini.e. node v _ito the bee-line that chlorophyll a calculates, λ _ifor node degree, be divided into in-degree and out-degree b _ifor node betweenness, α ⁺, α _-for weights, span is the real number of 0 to 1, and i represents node serial number, i=1,2 ..., 7;

Consider that average path length and cluster coefficients are on the impact of network, carry out second-order correction to the crucial degree model of node, correction model is:

${\mathrm{\&sigma;}}_i=(1-\mathrm{\&beta;}){\mathrm{\&sigma;}}_{i^{\&prime;}}+\mathrm{\&beta;}\frac{C_i\underset{j=1,j\&NotEqual;i}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&sigma;}}_{j^{\&prime;}}}{(n-1)L}$

In formula, C _ifor node v _icluster coefficients, represent node v _iwith the degree of coupling of other nodes in complex network, L is the average path length of network, j=1,2 ..., 7, β is weights, and span is the real number of 0 to 1;

Consider L, B _i, C _i, d _minand σ _ibe the key parameter that wawter bloom forms complex network characteristic, then build algal bloom and generate identification G parameter model:

$G=\exp (\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&sigma;}}_i^{\&prime;}\&times;d_{i\min })$

In formula, σ _i' representation node v _icrucial degree model σ _iwith node v _ithe product of design parameter value, i.e. σ _i'=c _i× σ _i, c _irepresent node v _idesign parameter value, n is nodes; Described concrete G parameter model comprises average path length L, node betweenness B _i, cluster coefficients C and bee-line d _min, and crucial degree model σ _i.