200923211 六、發明說明 【發明所屬之技術領域】 本發明係關於一雨水排放支援系統以及一雨水排放支 援方法，用以基於複數地區之目前時序降雨量或預測到之 降雨量，預測流入如泵廠或污水廠之一目標廠的進水流入 量，一雨水排放支援系統以及一雨水排放支援方法，用以 預測當下雨時流入如泵廠之一目標廠的流入水質，以及一 雨水排放控制系統與一雨水排放控制方法，用以控制在如 泵廠之一目標廠中的雨水泵。 【先前技術】 首先，係關於流入如泵廠之目標廠之進水流入量預測 之先前技術說明如下。 以雨水流進如泵廠或污水廠等污水排除之目標廠之流 入量預測方法而言，現今已經存在不少方法，例如以考量 地面鋪設狀況與污水管配置之實體模式爲基礎之RRL方 法（參閱已公開之日本專利編號322808/1994)、一種使用 神經網路的方法（參閱已公開之日本專利編號2000-257140) 以及以黑盒子模型爲基礎之方法，如像一種使用區塊導向 爲模型的方法（參閱已公開之日本專利編號2000-56835)。 根據以黑盒子爲模型之方法，該方法係以過去輸入/輸出 資料爲根據而預先構成的一流入量預測模型。在該流入量 預測模型中’藉由一雷達雨量計或複數之地面雨量計所量 測之降雨量作爲一輸入變數，以及一雨水流入量係作爲一 -4 - 200923211 輸出變數。透過流入量預測模型的使用’即可以現在降雨 量或預測降雨量爲基礎以預測一雨水流入量。 因爲雷達雨量計能夠精準地量測在一量測區域之降雨 強度，與藉由複數之地面雨量計所量測之降雨量相比’可 得到更多降雨量資訊的明細。因此，藉使用由雷達雨量計 所量測之降雨量或者以降雨量爲預測基礎之預測降雨量作 爲流入量預測模型如RRL方法或神經網路方法的一輸入 變數，則雨水流入一目標廠之流入量預測可以更精確的程 度來進行。 第二，係關於流入如泵廠之目標廠之流入水量與水質 的預測之先前技術說明如下。 在一混合之污水系統中，當雨下在一污水管所在的一 區域裡，則雨水一般係流進該污水管。位在該污水管之終 端處的一泵廠則排放包括流至一預定排放部位的雨水流入 水。 在這樣的污水下水道裡，當水係直接從一雨水泵井釋 放至一河川，應儘可能避免河川的污染。換言之，依照降 雨流入量與其水質，需要適當控制在一污水管裡水的收集 以及釋放到河川的水。在降雨啓動時，位在一污水管的污 染物質流出其流入的地方並流入一流入管，也就是發生一 般所稱之首次沖刷。首次沖刷的產生必須予以檢查並適當 地處置。 針對該類型之污水下水道，已提出的一發明從降雨量 來預測所流入水質，以便控制在消除污染物質成份的方法 -5- 200923211 運作量（參閱已公開之日本專利編號20 04-249200)。 第三，係關於在如栗廠之目標廠之雨水泵控制之先前 技術說明如下。 有一種知名的雨水泵控制器之控制方法，可事先決定 在該雨水泵的一啓動水位與一停止水位，當位在雨水泵井 之水位計所量測的値到達預定之一啓動水位或一停止水 位，則該控制器可控制該雨水泵的啓動或停止，其中該雨 水泵井係儲存流自於一污水管或其他之雨水。 當雨水突然流入雨水泵井時，爲了抑止雨水泵井之水 位上升，譬如由日本已公開之專利編號2000-3 2 8642所揭 露的一種預測控制器因運而生。在該預測控制器中’藉由 一地面雨量計或一雷達雨量計所量測的値可預測流到泵廠 的雨水量，以計算出用於排放水的雨水泵所需的運作數 量。該控制器可即時性變更雨水泵之啓動與停止水位。當 決定需要啓動一額外的水泵時’該雨水泵可在比預定水位 還要低的水位來啓動。 此外，在一混合污水系統的泵廠裡’將水的釋放係限 制在下面的方式，具有限制河川水質污染進入所釋放污水 之目的。換言之’當下一些雨時或者雨水流至栗廠之預測 流入量較小，雨水杲所做之水釋放可藉由設定雨水泵之啓 動水位與停止水位相對地高而予以抑止。 【發明內容】 首先，上述預測流進如泵廠之目標廠之進水流入量的 6 - 200923211 先前技術所遭遇問題係說明如下。一種使用神經網路的方 法（參閱已公開之日本專利編號2000-257140)以及以黑盒 子模型爲基礎之方法，當由一種黑盒子模型，如一種神經 網路的方法，所預估的流入量，藉最小平方的方法可從輸 入/輸出資料找出共同作用的參數，以構成流入量預測模 型。在本案例中，當該等輸入變數彼此高度相互關連時， 則不容易找出參數，因此很難構成一高精度之流入量預測 模型。當輸入至流入量預測模型的變數數目變大時，則不 易選擇合適的輸入變數以預測對目標廠之流入量，以及不 易分析與該選擇一致之流入量預測模型的構成。具體的 說’當降雨量係藉由一雷達雨量計所量測時，由於彼此間 具高度關連性的輸入變數數目明顯增加，則上述的問題係 易產生。 第二’上述預測流入如泵廠之目標廠之流入水質的先 前技術所遭遇的問題係說明如下。在已公開之日本專利編 號2004-249200所揭露預測流入水質之方法中，因爲該方 法係以水質模型方程式之公式表示，因此在本方法之方程 式較複雜。 第三’上述在如泵廠之目標廠之雨水栗控制之先前技 術所ia遇的問題係說明如下。透過地表流進污水管的雨， 藉流經該污水管以到達一泵廠，係複雜與廣泛的流程。因 此’在預測一流入量模型、預測一泵井水位模型，以及預 測雨水栗運作數目模型之中，降雨量之量測値係輸入至該 等模型’該等模型常產生錯誤性的預測値。因此，當真正 200923211 流入量小於預測量時’係造成過量的雨水泵在運作。另一 方面，當一真正流入量大於預測量時’係發生洪水。 本發明係在上述環境的觀點下予以開創出來。本發明 的一目標係提供雨水排放支援系統與雨水排放支援方法， 其能夠降低輸入至流入量預測單元之流入量預測模型的變 數數目，並促進藉由該流入量預測單元更精準預測雨水流 入量。 本發明的另一目標係提供以簡單與精確的方式在下雨 時預測進水品質之方法。本發明亦提供雨水排放支援系統 與雨水排放支援方法，其可藉使用由該預測方法預測到之 水質，或藉使用該預測到之水質與降雨期間之流入量，適 當地控制進水的蓄集與排出，以降低環境面的影響。 本發明的另外目標係，在當雨水流入量突然改變時在 用以變更雨水泵的啓動水位與停止水位以適當地排水的泵 控制器裡，提供雨水排放控制系統與雨水排放控制方法， 其可藉使用高度可靠性指標並配合流入量與預測到之水位 値’變更啓動水位與停止水位，以穩定運作一排放系統。 根據申請專利範圍第1項所述之一發明，係一雨水排 放支援系統’包含：一降雨量量測單元，用以量測複數地 區之降雨量；一線性對映單元，用以得到一降雨量時序資 料矩陣’該降雨量時序資料矩陣表示來自由該降雨量量測 單元在複數區域所量測到之降雨量之各地區之時序降雨 量’以及該降雨量時序資料矩陣執行一線性對映程序，以 將降雨量時序資料矩陣之多個變數資料轉換成較少之變數 -8- 200923211 資料而獲得線性對映資料矩陣；以及一流入量預測單元， 藉使用流入量預測模型以預測流入一目標廠之雨水流入 量，其中由該線性對映單元所得之線性對映資料矩陣係輸 入至該流入量預測模型。 根據申請專利範圍第7項所述之一發明，係一雨水排 放支援方法’用以預測流入一泵廠或一污水廠之雨水流入 量’該雨水排放支援方法包含以下步驟：藉由該降雨量量 測單元量測在複數區域之降雨量；藉由該線性對映單元， 得到一降雨量時序資料矩陣，該降雨量時序資料矩陣表示 來自由該降雨量量測單元在複數區域所量測到的降雨量之 各地區之時序降雨量，以及執行一線性對映程序，以將降 雨量時序資料矩陣之多個變數資料轉換成較少之變數資料 而獲得一線性對映資料矩陣；以及藉該流入量預測單元使 用一流入量預測模型，以預測雨水流入該目標廠之流入 量，其中由該線性對映單元所得之線性對映資料矩陣係輸 入至該流入量預測模型。 根據該雨水排放支援系統與該雨水排放支援方法，該 線性對映單元執行一線性對映程序，以將k X η之降雨 量時序資料矩陣轉換成一 k X m之線性對映資料矩陣 (m<n)。因此，可減少輸入至該流入量預測單元之流入量 預測模型之變數數目，以促進藉由該流入量預測單元對雨 水流入量更精準之預測。 根據申請專利範圍第2項所述之一發明，係一雨水排 放支援系統，包含：一降雨量預測單元，用以預測複數區 -9- 200923211 域在時序上之未來降雨量；一線性對映單元，用以得到一 預測到之降雨量時序資料矩陣，該預測到之降雨量時序資 料矩陣表示來自由該降雨量預測單元在複數區域所預測到 的未來降雨量之各地區之時序降雨量，以及執行一線性對 映程序，以將所預測到之降雨量時序資料矩陣之多個變數 資料轉換成較少之變數資料而獲得線性對映資料矩陣；以 及一流入量預測單元，藉使用流入量預測模型以預測流入 一目標廠之雨水流入量，其中由該線性對映單元所得之線 性對映資料矩陣係輸入至該流入量預測模型。 根據申請專利範圍第8項所述之一發明，係一雨水排 放支援方法，用以預測雨水流入一泵廠或一污水廠之流入 量，該雨水排放支援方法包含以下步驟：藉由該降雨量預 測單元在時序上預測複數區域之未來降雨量；藉由該線性 對映單元，得到預測降雨量時序資料矩陣，該矩陣表示來 自由該降雨量預測單元在複數區域所預測到的未來降雨量 之各地區之時序降雨量，以及執行一線性對映程序，以將 所預測到的降雨量時序資料矩陣之多個變數資料轉換成較 少之變數資料而使該線性對映單元獲得一線性對映資料矩 陣；以及藉該流入量預測單元使用一流入量預測模型，以 預測雨水流入該目標廠之流入量，其中由該線性對映單元 所得之線性對映資料矩陣係輸入至該流入量預測模型。 根據該雨水排放支援系統與該雨水排放支援方法，該 線性對映單元執行一線性對映程序，以將該k X η之預 測到的未來降雨量時序資料矩陣轉換成該k X m之線性 -10- 200923211 對映資料矩陣（m<n)。因此，可減少輸入至該流入量預測 單元之流入量預測模型之變數數目’以促進藉由該流入量 預測單元對雨水流入量更精準之預測。 在上述之雨水排放支援系統與雨水排放支援方法中， 更好作法係該線性對映單元藉使用一表示矩陣而執行該線 性對映程序，以獲得該線性對映資料矩陣，其中該表示矩 陣包含在過去時序資料矩陣之降雨量中之一變異矩陣或一 共變異矩陣的一固有向量元素。 根據上面之雨水排放支援系統與雨水排放支援方法， 用於該線性對映程序之表示矩陣係在過去降雨量時序資料 矩陣之一變異矩陣或一共變異矩陣之固有向量元素的組 合。因此，由該線性對映程序所得之線性對映資料矩陣元 素可由線性獨立時序資料組成，其中該等元素彼此間係無 關連性。 在上述之雨水排放支援系統與雨水排放支援方法中， 更好作法係以構成該表示矩陣之該固有向量的一固有値爲 基礎所計算出的一累積貢獻率大於一預定之臨限値。 根據雨水排放支援系統與雨水排放支援方法，形成該 表示矩陣使得以構成該表示矩陣之固有向量的固有値爲基 礎所計算出的該累積貢獻率，係大於一預定之臨限値。因 此，可得到該線性對映資料矩陣，以儘可能在遭受該線性 對映程序之前，防止該降雨量時序矩陣的資訊流失。 在上述之雨水排放支援系統與雨水排放支援方法中， 更好作法係該線性對映單元藉使用一表示矩陣而執行該線 -11 - 200923211 性對映程序，以獲得該線性對映資料矩陣，其中該表示矩 陣係由一負載矩陣所形成，而該負載矩陣係在分析過去降 雨量時序資料矩陣之一主要元素期間所得。 根據雨水排放支援系統與雨水排放支援方法，可得到 該線性對映資料矩陣以儘可能在遭受該線性對映程序之 前，防止該降雨量時序矩陣的資訊流失。此外，線性對映 資料矩陣的元素可由線性獨立時序資料組成，其中該等元 素彼此間係無關連性。 在上述之雨水排放支援系統與雨水排放支援方法中， 較佳的是該雨水排放支援系統另外包含一模型辨識單元， 用來以過去降雨量時序資料矩陣和過去降雨之時序流入量 爲基礎構成該流入量預測模型，其中該降雨量時序資料矩 陣之多個變數資料係被轉換成較少之變數資料。 根據雨水排放支援系統與雨水排放支援方法，可降低 輸入至該模型辨識單元之過去降雨量時序資料數目。因此 可促進該流入量預測模型的產生。另外，由於該流入量預 測模型係基於線性獨立降雨量時序資料所建構，因此可得 到一高精準度之流入量預測模型，其中在線性獨立降雨量 時序資料裡，該等元素彼此間係無關連性。 根據申請專利範圍第1 3項所述之一發明，係一雨水 排放支援方法’用以預測從一污水管流入一泵廠或一污水 廠之水質，該雨水排放支援方法包含以下步驟：以一預定 頻率量測該污水管所在一區域之降雨量；量測流入該泵廠 或污水廠之進水的水質；以及以現在水質、某些過去水 -12- 200923211 質、某些過去降雨量與某些過 由使用一非線性Hammer stein 預測一未來進水水質。 根據申請專利範圍第1 4 於一泵廠或一污水廠之一雨水 雨量量測措施，用以量測該污 一水質量測措施，用以量測流 質；一資料獲取與記憶措施， 雨量量測措施與該水質量測措 水質；一預測措施，以由該資 現在水質、某些過去水質、某 雨量之指數爲基礎，藉由使用 的一系統辨識方法，以預測一 支援措施，藉使用以該預測措 以產生與蓄集和排出流入污水 池之進水相關一運作命令。 在上述之雨水排放支援系 施以由資料獲取與記憶措施所 水質、某些過去降雨量與某些 藉由使用非線性H a m m e r s t e i η 該未來水質還有進水之一流入 藉使用由該預測裝置所預測到 入量，產生與蓄集和排出流入 滯留池之進水相關的運作命令 去降雨量之指數爲基礎，藉 模型的一系統辨識方法，以 項所述之一發明，係提供用 排放支援系統，包含：一降 水管所在一區域之降雨量； 入該泵廠或污水廠的進水品 用以獲取與記憶分別由該降 施所定期量測到之降雨量與 料獲取與記憶措施所提供之 些過去降雨量與某些過去降 —非線性Hammer stein模型 未來進水水質；以及一運作 施所預測到之未來水質，用 泵井、雨水泵井或雨水滯留 統中，更好作法係該預測措 提供之現在水質、某些過去 過去降雨量之指數爲基礎， 模型的系統辨識方法，預測 量，以及該運作支援措施係 之該未來水質還有進水之流 污水泵井、雨水泵井或雨水 -13- 200923211 在上述之雨水排放支援系統中，更好作法係將一第二 水質量測措施設於該污水管中，該資料獲取與記憶措施則 進一步定期取得與記憶由該第二水質量測措施在該污水管 中所量測到之水質，該預測措施藉由使用非線性 Hammer stein模型的系統辨識方法，以該污水管中的現在 水質、該污水管中的某些過去水質、某些過去降雨量與某 些過去降雨量之指數爲基礎，預測在流入該泵廠或污水廠 之前該污水管中的未來水質，以及該運作支援措施藉進一 步使用流入該栗廠或污水廠之前該污水管中的未來水質， 產生與蓄集和排出流入該污水泵井、該雨水泵井或該雨水 滯留池之進水相關的該運作命令。 在上述之雨水排放支援系統中，更好作法係將一第二 水量量測措施設於該污水管中，該資料獲取與記憶措施則 進一步定期取得與記憶由該第二水量量測措施在該污水管 中所量測到之水流量，以及該預測措施藉由使用非線性 Hammer stein模型的系統辨識方法，以該污水管中的現在 水流量、該污水管中的某些過去水流量、某些過去降雨量 與某些過去降雨量之指數爲基礎，預測在流入該泵廠或污 水廠之前該污水管中的未來水流量，以及該運作支援措施 藉進一步使用流入該泵廠或污水廠之前該污水管中的未來 水流量，產生與蓄集和排出流入該污水泵井、該雨水泵井 或該雨水滯留池之進水相關的該運作命令。 在上述之雨水排放支援系統中，更好作法係將一水位 量測措施設於該污水管中，該資料獲取與記憶裝置則進一 -14- 200923211 步定期取得與記憶由該水位量測措施在該污水管中所量測 到之水位，該資料獲取與記憶措施進一步得到來自一氣象 資訊系統之氣象資訊，以及該預測措施可得到來自該資料 獲取與記憶裝置在一晴天時於該污水管中之水位，以計算 於該污水管中之污泥或沈積物之位準，並基於在一雨天時 於該污水管中之水流量與水質和上述在該污水管中的已算 出之污泥或沈積物位準，預測流入該泵廠或污水廠之未來 水流量與水質。 在上述之雨水排放支援系統中，更好作法係當所量測 到之進水水質比預定臨限値還要差時，該運作支援措施產 生該運作命令以使設於該污水泵井中之污水泵運作，同時 除非該降雨量大於一上臨限値，否則暫停設於該雨水泵井 中之雨水泵。 在上述之雨水排放支援系統中，更好作法係該運作支 援措施自預測到之進水流入量辨別出一首次沖刷，以及產 生運作命令以使該首次沖刷之水匯聚在該雨水滯留池中， 並當降雨量低於一預定値時，使設於該雨水滞留池中之泵 將上述匯聚之水送回至污水栗井。 根據申請專利範圍第2 1項所述之一發明，係一雨水 排放控制系統，用於以流入一泵廠之雨水流入量爲基礎決 定雨水泵的運作數目，以及以該泵廠中之水位變化爲基礎 控制該雨水泵之啓動與停止，該雨水排放控制系統包含： 一偵測措施，用以偵測在一預定上游地點之水位，其中水 係流到該泵廠；以及一變更措施，用以基於符合該偵測措 -15- 200923211 施所偵測到之水位的一水流狀況，變更至少該等雨水泵之 一預定啓動水位與一預定停止水位的其中之一。 在上述之雨水排放控制系統中，更好作法係當該預定 地點之水位高於一預定臨限値時，則該變更措施變更至少 該等雨水泵之該啓動水位與該停止水位的其中之一。 在上述之雨水排放控制系統中，更好作法係當該預定 地點之水位低於一預定臨限値時，則該變更措施變更至少 該等雨水泵之該啓動水位與該停止水位的其中之一。 在上述之雨水排放控制系統中，更好作法係在該預定 地點之水位以及在該預定地點之水位變化速度兩者皆作爲 在該預定地點之一水位指標。 在上述之雨水排放控制系統中，更好作法係該雨水排 放控制系統另外包含一預測措施，用以預測流入該泵廠之 雨水流入量，以及一計算措施，用以計算雨水栗的一運作 數目’該等雨水泵需排出該預測措施所預測到之雨水流入 量之水。 根據申請專利範圍第2 6項所述之一發明，係一雨水 排放控制方法，包含以下步驟：基於流入一泵廠之一預測 到的雨水流入量’決定雨水泵的一運作數目；基於該泵廠 之一水位改變’控制該等雨水泵的啓動與停止；以及基於 靠近該栗廠之一預定上游地點之一污水管中的一水位，變 更該等雨水泵的一啓動水位與一停止水位，以延遲該等雨 水泵之一啓動時間。 根據申請專利範圍第2 7項所述之一發明，係一雨水 -16- 200923211 排放支援方法，用以預測從一污水管流入一栗廠或污水廠 之水流入量，該雨水排放支援方法包含以下步驟：以一預 定頻率量測該污水管所在一區域之降雨量；量測流入該泵 廠或污水廠之水流量；以及藉由使用一非線性 Hammerstein模型的一系統辨識方法，以現在流量、某些 過去流量、某些過去降雨量與某些過去降雨量之指數爲基 礎，以預測未來水流入量。 根據申請專利範圍第2 8項所述之一發明，係一雨水 排放控制系統，用以藉由使用申請專利範圍第7項所述之 雨水排放支援方法所預測到之雨水流入量，基於流入一泵 廠之一雨水流入量，決定雨水泵的一運作數目，以及基於 該泵廠中之一水位改變，控制該等雨水泵的啓動與停止， 該雨水排放控制系統包含：一偵測措施，用以偵測在一預 定上游地點之一水位，其中水係流到該泵廠；一變更措 施，用以基於符合該偵測措施所偵測到之水位的流動狀 況，變更至少該等雨水泵之一預定啓動水位與一預定停止 水位的其中之一；以及一計算措施，用以計算雨水泵的一 運作數目，該等雨水泵需排出申請專利範圍第7項所述之 雨水排放支援方法所預測到之雨水流入量之水。 根據本發明之雨水排放支援系統與雨水排放支援方 法，該線性對映單元係執行一線性對映程序，將該k X η 矩陣之降雨量時序資料矩陣或該k χ η矩陣之預測降雨 時序資料矩陣予以轉換成一 k X m矩陣之線性對映資料 矩陣（m<n)。因此，可減少輸入至該流入量預測單元之流 -17- 200923211 入量預測模型之變數數目，以促進藉由該流入量預測單元 對雨水流入量更精準之預測。 根據本發明之另一雨水排放支援系統與雨水排放支援 方法’由於進水水質係由基於進水水質之量測値與過去降 雨量之系統辨識方法（非線性Hammerstein模型）所預測， 因此可容易執行該預測。因爲進水的蓄集與排出可適當控 制而與預測結果保持一致性，因此減少在環境面的不良效 應。 根據雨水排放支援系統與雨水排放支援方法，雨水泵 之一啓動水位與一停止水位係基於流入一泵廠之一預測流 入量所變更，符合一水流狀況。因此可穩定運作一排放系 統。 【實施方式】 圖1係一描述在本發明之一實施例之雨水排放系統之 架構圖。圖2係一描述在如圖1所示之雨水排放系統之一 控制系統所組成之方塊圖。 如圖1所示，在本實施例之雨水排放系統包括一降雨 量量測單元1 0，用以量測一污水管2 0或其他同等物所在 一區域之複數地區之降雨量。包括一流入管21與一雨水 泵井30之一泵廠係座落在污水管20的下游。 降雨量量測單元1 〇係由一雷達雨量計11與複數之地 面雨量計1 2所組成。在本案例中，降雨量量測單元1 〇係 由該雷達雨量計11所構成，一雷達基地台以η片區塊來 -18- 200923211 整體量測一降雨量’利用輻射波的直線傳播特性以及當與 雨滴接觸所發生相同的反射特性。如圖1所示，由一觀測 網所量測降雨量之一區域係分成η片區塊，使得各區塊係 由一方形所組成’方形的每邊係有數百公尺到數公里之 長。因此所分成η片區塊的降雨量可分別被量測。 當降雨量量測單兀1 0係由雷達雨量計1 1所構成時， 比起當降雨量量測單元1 0係由複數之地面雨量計1 2所組 成時，則所量測降雨量的一區域可細分成一較小觀測網。 因此，可增加由該地面雨量計所量測之該等區域數目η。 例如污水管20係一混合型態。落在污水管20所在一 區域之雨水，加上在此區域之家用廢水與工業廢水共同流 經污水管2 0。 有一泵廠座落在污水管2 0的下游，一流入管2 1係設 於該廠。流入管2 1暫時收集從污水管2 0進來之雨水。如 圖1所示，一水質量測計2 5、一流速計2 6，以及一水位 計27係位在污水管20且靠近流入管21。 在泵廠裡，淤沙沈澱池之雨水泵井3 0係座落於流入 管2 1的下游。該泵廠亦設有污水泵井3 1與雨水滯留池 32。進水透過流入管21與淤沙沈澱池（圖中並沒表示）， 自污水管2 0流至淤沙沈澱池之雨水泵井3 0、污水泵井3 1 與雨水滯留池3 2。淤沙沈澱池之雨水泵井3 0通常在流入 管2 1的下游具有溢流堰。經常性之進水流入污水泵井 3 1。在另一方面，當包括雨水的大量進水在降雨期間流入 淤沙沈澱池之雨水泵井3 0時，則流入管2 1的水位上升。 -19- 200923211 因此，溢過高溢流堰的水即流入淤沙沈澱池之雨水栗 3 0。故雨水滯留池3 2匯聚該處的水流入泵廠。雨水滯 池3 2透過圖中並沒表示之溢流堰接收流入其中之進水 此處並沒表示之一溢流堰係座落於雨水滞留池3 2與流 管2 1之間，或此處並沒表示之一閘門係座落於雨水滯 池32與流入管21之間。 淤沙沈澱池之雨水泵井3 0係包括多個雨水泵3 5。 入泥沙沈澱池雨水栗井3 0之進水（雨水）係藉由該雨水 35釋放至一河川34。污水泵井31包括一污水泵36。 入污水泵井3 1之進水（污水）係於此處匯集，並透過污 泵36送至一污水處置設備，圖中並沒表示，污水泵36 受制於一污水處置。接著被處置過的水即釋放至一河川 其他相等物。污水若超出污水處置設備所能處理的範_ 則將污水送至雨水滯留池3 2以匯聚其中，以避免該污 溢出污水泵井3 1。雨水滯留池3 2包括一雨水滯留池 3 7。當流進流入管2 1的水明顯減少時，則匯聚於雨水 留池3 2裡的水係透過雨水滯留池泵3 7被送回至污水衆 3 1° 換言之，從污水管2 0流出之進水，透過位於污水 2〇的一終點接頭與淤沙沈澱池之間的流入管2 1，圖中 沒表示’匯聚於淤沙沈澱池之雨水泵井3 0、污水栗井 與雨水滞留池32的其中之一或者全部，然後該進水被 放至一預定排放目的地。 流速計2 3與水質量測計2 4係座落於流入管21。 井 留 入 留 流 栗 流 水 係 或 水 泵 滯 井 管 並 31 釋 流 -20- 200923211 速計2 3係量測匯集於流入管2 1的進水水位。水 24係量測流進流入管2 1的進水水質。 如圖1所示，淤沙沈澱池之雨水泵井3 0包 計3 8以量測淤沙沈澱池之雨水泵井3 0的水位。 到位在流入管2 1之流速計23所量得的値以及位 澱池之雨水栗井3 0之流速計3 8所量得的値。 雨水排放系統的一控制系統係以圖說明之。 如圖2所示，雨水排放系統包括：一流入量 40，藉使用在複數地區之現在時序降雨量，以預 泵廠之一目標廠的進水流入量；一進水水質] 5 0，藉使用上述之降雨量，以預測流入如泵廠之 (如流入管2 1 )的進水水質；以及一雨水泵控制損 藉使用流入量預測機制40所預測的進水流入量 至少該等雨水泵3 5之一啓動水位與一停止水位 -* 〇 流入量預測機制40之細部架構，參閱圖3 j 說明。 如圖3所示，流入量預測機制40係設有一 單元4 1與流入量預測單元42，其中線性對映單 施行一線性對映過程，該線性對映過程係轉換降 單元1 〇所量測之複數地區之降雨量時序資料矩 流入量預測單元42藉使用由線性對映單元4 1所 對映資料，以預測流入一目標區域之雨水流入量 陣製作單元43係連接至線性對映單元4 1，用以 質量測計 括一流速 隨時可得 在淤沙沈 預測機制 測流入如 頁測機制 一目標廠 !制 6 0， ，以調整 的其中之 缚圖4的 線性對映 元41係 雨量量測 陣，以及 得之線性 。表示矩 製作一表 -21 - 200923211 不矩陣’提供給線性對映單元41作爲線性對映過程。模 型辨識單元44係連接至流入量預測單元42 ,用以組成一 流入量預測模型，提供給流入量預測單元4 2作爲雨水流 入量之預測。 線性對映單元41得到一降雨量時序資料矩陣X，其 中該降雨量時序資料矩陣X係表示各地區之時序降雨 量，而各地區之時序降雨量係來自降雨量量測單元1 〇所 量測複數地區之降雨量，以及線性對映單元41施行一線 性對映過程，將降雨量時序資料矩陣X之多個變數轉換 成較少變數之資料，以得到一線性對映資料矩陣Y。 具體而言，線性對映單元4 1係執行下面方程式（1)的 算術運算。 γ = XP……方程式（1) (其中X係一降雨量時序資 料 k(離散時間）X n(所量測降雨之區塊數）之矩陣，與P 係一 η X m矩陣（m<n)之表示矩陣，該表示矩陣係由表 示矩陣製作單元43所產生，說明如下） 在降雨量時序資料矩陣中，在時間t各區塊降雨量相 當於一列向量Xt (t = 1，…，k)。其中，目前最近之降雨量 資料係疋義成Xi ’ q個步驟（lSqgk-1)之則的降雨量資 料係定義成xq +丨。 線性對映單元4 1係執行方程式（1 )計算而得到k X m 矩陣之線性對映資料矩陣Y。 流入量預測單元4 2使用由下述模型辨識單元4 4所組 成之流入量預測模型，以預測雨水流入量’其中由該線性 -22- 200923211 對映單元41所得之線性對映資料係輸入至該流入量預測 模型。如圖3所示，當流入量預測單元42正預測雨水流 入量時，則應將該時序流入量資料納入考量。 其中，流入量預測模型係一黑盒子模型，該模型係由 過去所量測之降雨量資料與相當於該降雨量資料之流入量 資料兩者的關係所決定。在考慮到由線性對映單元41所 得之到目前爲止之降雨量線性對映資料以及到目前爲止之 時序流入量資料’則流入量預測模型已依照降雨量資料與 過去流入量資料的關係而事先製作出來，故流入量預測單 元42可根據該流入量預測模型以預測雨水之未來流入 量。 表示矩陣製作單元43藉由多變數分析之一的主要分 量分析以計算來自過去降雨量時序資料的一主要元素，而 得到一負載矩陣，以及藉由所得之負載矩陣的特性，將該 矩陣轉移成表不矩陣P，該表示矩陣P係作爲線性對映單 元41之線性對映過程。具體而言，η X m矩陣之表示矩 陣P係由過去降雨量時序資料矩陣之變異矩陣或共同變異 矩陣之多個固有向量元素組合所製作出來。 參閱圖4以詳盡說明表示矩陣製作單元43。 在步驟1 1中’從過去降雨量時序資料而得以計算出 變異矩陣或共同變異矩陣’其中該過去降雨量時序資料係 在一區塊的一 η地區所量測之降雨量。當過去降雨量時序 資料係以k h X η的矩陣X h方式表示時，構成該表示矩陣 P的元素係爲矩陣Xh之變異矩陣或共同變異矩陣Sh的一 -23- 200923211 η階固有向量Pj(j = 1，...，n)之行向量ρι，…，Pm(m<n)。注意 關於固有向量Pj，係使用滿足；I 又2$…2 λ …入η 狀況之一固有値Λ j (；j = 1 , ...，η)。 在步驟1 1中，如下面方程式（2)所示’第一主要分量 到第m主要分量在離散時間t係表示成kh X m的矩陣T 的一 m階之列向量ti(i = 1，...，k)。200923211 VI. Description of the Invention [Technical Fields of the Invention] The present invention relates to a rainwater drainage support system and a rainwater drainage support method for predicting inflows such as pump plants based on current time series rainfall or predicted rainfall in a plurality of regions. Or the influent inflow of a target plant in the WWTP, a rainwater drainage support system and a rainwater drainage support method to predict the influx of water flowing into a target plant such as a pumping plant when it rains, and a rainwater discharge control system and A rainwater discharge control method for controlling a rainwater pump in a target plant such as a pump factory. [Prior Art] First, the prior art description of the inflow of inflow into a target plant such as a pump factory is as follows. In terms of the inflow forecast method of rainwater flowing into a target plant such as a pump plant or a sewage plant, there are already many methods, such as the RRL method based on the physical model of the ground laying condition and the sewage pipe configuration ( See Japanese Patent No. 322808/1994, a method using a neural network (refer to the published Japanese Patent No. 2000-257140), and a method based on a black box model, such as a model using block orientation. Method (refer to the published Japanese Patent No. 2000-56835). According to the method of modeling a black box, the method is an inflow prediction model which is preliminarily constructed based on past input/output data. In the inflow prediction model, the rainfall measured by a radar rain gauge or a plurality of ground rain gauges is used as an input variable, and a rainwater inflow is used as a -4 - 200923211 output variable. The use of the inflow forecasting model can be based on current rainfall or predicted rainfall to predict a rainwater inflow. Because the radar rain gauge can accurately measure the rainfall intensity in a measurement area, the rainfall information can be obtained in comparison with the rainfall measured by a plurality of ground rain gauges. Therefore, by using the rainfall measured by the radar rain gauge or the predicted rainfall based on the rainfall as an input variable of the inflow prediction model such as the RRL method or the neural network method, the inflow of rainwater into a target plant Volume prediction can be done to a more precise extent. Second, the prior art description of the inflow of water and water quality into the target plant such as the pump plant is as follows. In a mixed sewage system, when rain falls in an area where a sewage pipe is located, rainwater generally flows into the sewage pipe. A pump plant located at the end of the sewer pipe discharges rainwater influx including a predetermined discharge site. In such sewage sewers, when the water system is directly released from a rain pump well to a river, the pollution of the river should be avoided as much as possible. In other words, depending on the amount of rain inflow and its water quality, it is necessary to properly control the collection of water in a sewer pipe and the release of water into the river. At the start of the rain, the pollutants in a sewage pipe flow out of the inflow and into an inflow pipe, which is what is commonly referred to as the first flush. The production of the first flush must be checked and disposed of properly. For this type of sewage sewer, an invention has been proposed to predict the inflow of water from rainfall in order to control the amount of pollutants to be removed - 5, 2009,232,11 (see Japanese Patent No. 20 04-249200). Third, the previous technical description of the rain pump control at the target plant in Ruli Factory is as follows. There is a well-known control method of the rainwater pump controller, which can determine in advance a starting water level of the rainwater pump and a stop water level, when the water level gauge measured in the rainwater pump well reaches a predetermined starting water level or one When the water level is stopped, the controller can control the start or stop of the rainwater pump, wherein the rainwater pump well is stored from a sewage pipe or other rainwater. When the rainwater suddenly flows into the rainwater pump well, in order to suppress the rise of the water level of the rainwater pump well, a predictive controller disclosed in Japanese Patent Publication No. 2000-3 2 8642 is created. In the predictive controller, the amount of rain flowing to the pumping plant can be predicted by a ground rain gauge or a radar rain gauge to calculate the amount of operation required for the rainwater pump for draining water. The controller can change the start and stop water levels of the rain pump in real time. When it is decided that an additional pump needs to be started, the rain pump can be started at a lower water level than the predetermined water level. In addition, in the pumping plant of a mixed sewage system, the release of water is limited to the following method, which has the purpose of restricting the pollution of river water into the discharged sewage. In other words, when some rain or rainwater flows to the chestnut factory, the predicted inflow is small, and the water release from the rainwater can be suppressed by setting the start water level of the rainwater pump to be relatively high. SUMMARY OF THE INVENTION First, the above-mentioned problems predicted by the prior art flowing into the target plant of the pump factory are explained as follows. A method of using a neural network (refer to the published Japanese Patent No. 2000-257140) and a method based on a black box model, when the inflow is estimated by a black box model, such as a neural network method. The least squares method can be used to find the interaction parameters from the input/output data to form the inflow prediction model. In this case, when the input variables are highly correlated with each other, it is not easy to find the parameters, so it is difficult to construct a high-precision inflow prediction model. When the number of variables input to the inflow prediction model becomes large, it is not easy to select an appropriate input variable to predict the inflow to the target plant, and it is not easy to analyze the composition of the inflow prediction model consistent with the selection. Specifically, when the rainfall is measured by a radar rain gauge, the above problems are easily generated due to a significant increase in the number of input variables having a high degree of correlation with each other. The second 'previous problems encountered in predicting the inflow of water into the target plant such as the pump plant are described below. In the method for predicting the inflow of water as disclosed in Japanese Laid-Open Patent Publication No. 2004-249200, since the method is expressed by the formula of the water quality model equation, the equation in the method is complicated. The third problem mentioned above in the prior art of rainwater chestnut control at the target plant of the pump factory is as follows. Rain flowing through the surface into the sewer pipe flows through the sewer pipe to reach a pump plant, which is a complex and extensive process. Therefore, in predicting an inflow model, predicting a pumping well water level model, and predicting the number of rainwater chestnut operations, the amount of rainfall is measured into the models. These models often produce erroneous predictions. Therefore, when the actual 200923211 inflow is less than the predicted amount, the excess rainwater pump is operating. On the other hand, when a true inflow is greater than the predicted amount, a flood occurs. The present invention has been developed in view of the above circumstances. An object of the present invention is to provide a rainwater drainage support system and a rainwater drainage support method capable of reducing the number of variables of an inflow prediction model input to an inflow amount prediction unit, and facilitating more accurate prediction of rainwater inflow by the inflow prediction unit . Another object of the present invention is to provide a method for predicting the quality of incoming water when it is raining in a simple and precise manner. The present invention also provides a rainwater drainage support system and a rainwater drainage support method, which can appropriately control the accumulation of influent water by using the water quality predicted by the prediction method or by using the predicted water quality and the inflow amount during the rainfall period. And discharge to reduce the impact of the environmental surface. Another object of the present invention is to provide a rainwater discharge control system and a rainwater discharge control method in a pump controller for changing the start water level of the rainwater pump and stopping the water level to properly drain when the rainwater inflow suddenly changes. By using the high reliability index and matching the inflow and predicted water level 値 'change the starting water level and stop the water level to stabilize the operation of an emission system. According to one of the inventions mentioned in claim 1, a rainwater drainage support system includes: a rainfall measuring unit for measuring rainfall in a plurality of regions; and a linear mapping unit for obtaining a rainfall The time series data matrix 'the rainfall time series data matrix represents the time series rainfall from each region of the rainfall measured by the rainfall measuring unit in the plurality of regions' and the rainfall time series data matrix performs a linear mapping a program for converting a plurality of variable data of a rainfall time series data matrix into a variable -8-200923211 data to obtain a linear mapping data matrix; and an inflow prediction unit by using an inflow prediction model to predict an inflow one The rainwater inflow of the target plant, wherein the linear mapping data matrix obtained by the linear mapping unit is input to the inflow prediction model. According to one of the inventions described in claim 7, a rainwater discharge support method is used to predict the amount of rainwater inflow into a pump plant or a sewage plant. The rainwater drainage support method comprises the following steps: The measuring unit measures the rainfall in the complex region; by the linear mapping unit, a rainfall time series data matrix is obtained, and the rainfall time series data matrix is obtained from the plurality of regions measured by the rainfall measuring unit Time-series rainfall for each region of the rainfall, and the implementation of a linear mapping procedure to convert a plurality of variable data of the rainfall timing data matrix into less variable data to obtain a linear mapping data matrix; The inflow prediction unit uses an inflow prediction model to predict the inflow of rainwater into the target plant, wherein the linear mapping data matrix obtained by the linear mapping unit is input to the inflow prediction model. According to the rainwater drainage support system and the rainwater drainage support method, the linear mapping unit performs a linear mapping process to convert the k X η rainfall quantity time series data matrix into a k X m linear mapping data matrix (m <n). Therefore, the number of variables input to the inflow amount prediction model of the inflow amount prediction unit can be reduced to promote a more accurate prediction of the rainwater inflow amount by the inflow amount prediction unit. According to one of the inventions mentioned in claim 2, the invention relates to a rainwater drainage support system, comprising: a rainfall prediction unit for predicting the future rainfall of the complex region-9-200923211 domain in time series; a linear mapping a unit for obtaining a predicted rainfall time series data matrix, the predicted rainfall time series data matrix indicating time series rainfall from each region of the future rainfall predicted by the rainfall prediction unit in the plurality of regions, And performing a linear mapping process to convert the plurality of variable data of the predicted rainfall time series data matrix into a smaller number of variable data to obtain a linear mapping data matrix; and an inflow prediction unit by using the inflow amount The model is predicted to predict the amount of rainwater inflow into a target plant, wherein the linear mapping matrix obtained by the linear mapping unit is input to the inflow prediction model. According to one of the inventions described in claim 8, a rainwater discharge support method for predicting the inflow of rainwater into a pump plant or a sewage plant includes the following steps: The prediction unit predicts the future rainfall of the complex region in time series; by using the linear mapping unit, a predicted rainfall time series data matrix is obtained, which represents the future rainfall predicted from the plurality of regions by the rainfall prediction unit. Time-series rainfall for each region, and performing a linear mapping procedure to convert a plurality of variable data of the predicted rainfall timing data matrix into less variable data to obtain a linear mapping of the linear mapping unit a data matrix; and an inflow prediction model using the inflow prediction unit to predict the inflow of rainwater into the target plant, wherein the linear mapping matrix obtained by the linear mapping unit is input to the inflow prediction model . According to the rainwater drainage support system and the rainwater drainage support method, the linear mapping unit performs a linear mapping process to convert the predicted future rainfall timing data matrix of k X η into the linearity of the k X m - 10- 200923211 Mapping data matrix (m <n). Therefore, the number of variables of the inflow amount prediction model input to the inflow amount prediction unit can be reduced to promote a more accurate prediction of the rainwater inflow amount by the inflow amount prediction unit. In the above-described rainwater drainage support system and rainwater drainage support method, it is a better practice for the linear mapping unit to perform the linear mapping process by using a representation matrix to obtain the linear mapping data matrix, wherein the representation matrix comprises An inherent vector element of a variation matrix or a common variation matrix in the rainfall of the temporal data matrix. According to the above rainwater drainage support system and the rainwater drainage support method, the representation matrix for the linear mapping program is a combination of one of the past rainfall time series data matrix variation matrix or the common vector element of the common variance matrix. Therefore, the linear mapping data matrix elements obtained by the linear mapping procedure can be composed of linear independent time series data, wherein the elements are not related to each other. In the above-described rainwater drainage support system and rainwater drainage support method, a better practice is to calculate a cumulative contribution rate based on an inherent enthalpy of the eigenvectors constituting the representation matrix, which is greater than a predetermined threshold 値. According to the rainwater drainage support system and the rainwater drainage support method, the representation matrix is formed such that the cumulative contribution rate calculated based on the inherent enthalpy of the eigenvectors constituting the representation matrix is greater than a predetermined threshold 値. Therefore, the linear mapping data matrix can be obtained to prevent information loss of the rainfall timing matrix as much as possible before suffering the linear mapping procedure. In the above-described rainwater drainage support system and rainwater drainage support method, it is a better practice for the linear mapping unit to perform the line-11-200923211 sexual mapping program by using a representation matrix to obtain the linear mapping data matrix. Wherein the representation matrix is formed by a load matrix obtained during analysis of one of the main elements of the past rainfall timing data matrix. According to the rainwater drainage support system and the rainwater drainage support method, the linear mapping data matrix can be obtained to prevent the information loss of the rainfall timing matrix from being subjected to the linear mapping procedure as much as possible. In addition, the elements of the linear mapping data matrix may be composed of linear independent time series data, wherein the elements are independent of one another. In the above-described rainwater drainage support system and rainwater drainage support method, it is preferable that the rainwater drainage support system further includes a model identification unit for constructing the basis based on the past rainfall time series data matrix and the temporal inflow of past rainfall. The inflow prediction model, wherein the plurality of variable data of the rainfall time series data matrix is converted into less variable data. According to the rainwater drainage support system and the rainwater drainage support method, the number of past rainfall timing data input to the model identification unit can be reduced. Therefore, the generation of the inflow prediction model can be promoted. In addition, since the inflow prediction model is constructed based on linear independent rainfall time series data, a high-precision inflow prediction model can be obtained, in which the elements are independent of each other in the linear independent rainfall time series data. Sex. According to one of the inventions described in claim 13 of the patent application, a rainwater discharge support method is used to predict the water quality flowing from a sewage pipe to a pump factory or a sewage plant, and the rainwater discharge support method comprises the following steps: Predetermined frequency to measure the rainfall in an area where the sewer pipe is located; measure the water quality of the influent water flowing into the pump or sewage plant; and use the current water quality, some past waters -12-200923211, some past rainfall and Some have used a nonlinear Hammer stein to predict a future influent water quality. According to the application of the patent scope, the rainwater and rainfall measurement measures of a pump factory or a sewage treatment plant are used to measure the pollution-water quality measurement measures for measuring fluid quality; a data acquisition and memory measure, rainfall quantity The measurement measures and the quality of the water quality measure; a forecasting measure based on the current water quality of the fund, some past water quality, and a certain rainfall index, by using a system identification method to predict a support measure, by using This prediction is used to generate an operational command associated with the accumulation and discharge of the incoming water into the lagoon. In the above-mentioned rainwater drainage support system, the water quality from the data acquisition and memory measures, some past rainfall and some use of the non-linear H ammerstei η, the future water quality and one of the influent flows are used by the prediction device. The predicted intake amount is generated based on the index of the operation order related to the water inflow and discharge into the detention tank, and the system is provided by one of the inventions. The system includes: rainfall in a region where a downcomer is located; water intake into the pump or sewage plant for obtaining and memorizing the rainfall and material acquisition and memory measures regularly measured by the reduced plant Providing some past rainfall and some past declines – the future influent water quality of the nonlinear Hammer stein model; and the future water quality predicted by the operational plant, using pump wells, rainwater pump wells or rainwater retention systems, better practice The forecast provides the basis for the current water quality, some past past rainfall indices, the model's system identification method, the forecasted amount, and the operation. The future water quality of the aid measures is also the influent sewage pump well, rain pump well or rainwater-13- 200923211 In the above-mentioned rainwater drainage support system, it is better to set a second water quality measurement measure in the In the sewage pipe, the data acquisition and memory measures are further regularly obtained and memorized by the water quality measured by the second water quality measurement measure in the sewage pipe. The prediction measures are identified by the system using a nonlinear Hammerstein model. Method, based on the current water quality in the sewer, some past water quality in the sewer, some past rainfall and some past rainfall indices, predicting the sewer before entering the pump or sewage plant The future water quality in the medium and the operational support measures to generate and accumulate and discharge into the sewage pump well, the rainwater pump well or the rainwater retention tank by further using the future water quality in the sewage pipe before flowing into the chestnut or sewage plant The operational order related to the influent water. In the above-mentioned rainwater drainage support system, it is better to set a second water quantity measurement measure in the sewage pipe, and the data acquisition and memory measures are further periodically obtained and memorized by the second water quantity measurement measure. The water flow measured in the sewer pipe, and the prediction method by the system identification method using the nonlinear Hammerstein model, the current water flow in the sewage pipe, some past water flow in the sewage pipe, some Based on past rainfall and some index of past rainfall, predicting future water flows in the sewer before entering the pump or wastewater plant, and the operational support measures before further use into the pump or wastewater plant The future water flow in the sewer pipe produces the operational command associated with accumulating and discharging the influent water flowing into the sewage pump well, the rainwater pump well or the rainwater retention tank. In the above-mentioned rainwater drainage support system, it is better to set a water level measurement measure in the sewage pipe. The data acquisition and memory device is further obtained and memory is measured by the water level measurement measure. The water level measured in the sewage pipe, the data acquisition and memory measures further obtain meteorological information from a meteorological information system, and the prediction measure can be obtained from the data acquisition and memory device in the sewage pipe on a sunny day. The water level is calculated based on the level of sludge or sediment in the sewage pipe and based on the water flow and water quality in the sewage pipe on a rainy day and the calculated sludge or the above-mentioned sludge in the sewage pipe or The sediment level is predicted to predict future water flow and water quality flowing into the pump or wastewater plant. In the above-mentioned rainwater drainage support system, it is better to use the operation support measure to generate the sewage in the sewage pump well when the measured inlet water quality is worse than the predetermined threshold. The pump operates, and unless the rainfall is greater than the upper limit, the rain pump installed in the rain pump well is suspended. In the above-mentioned rainwater drainage support system, it is a better practice for the operational support measure to discriminate a first flush from the predicted influent inflow, and to generate an operational command to concentrate the first flushed water in the rainwater retention tank, And when the rainfall is lower than a predetermined threshold, the pump disposed in the rainwater retention tank returns the concentrated water to the sewage chestnut. According to one of the inventions described in claim 21, a rainwater discharge control system is used to determine the number of rainwater pump operations based on the amount of rainwater flowing into a pump plant, and to vary the water level in the pump plant. Controlling the start and stop of the rainwater pump based on the following: a detection measure for detecting a water level at a predetermined upstream location, wherein the water system flows to the pump factory; and a change measure, And changing at least one of a predetermined starting water level and a predetermined stopping water level of one of the rainwater pumps based on a water flow condition that is consistent with the water level detected by the detecting means -15-200923211. In the above-described rainwater discharge control system, it is a better practice to change at least one of the starting water level of the rainwater pump and the stop water level when the water level of the predetermined location is higher than a predetermined threshold. . In the above-described rainwater discharge control system, a better practice is to change at least one of the starting water level of the rainwater pump and the stop water level when the water level of the predetermined location is below a predetermined threshold. . In the above-described rainwater discharge control system, it is better to use both the water level at the predetermined location and the water level change velocity at the predetermined location as one of the water level indicators at the predetermined location. In the above-mentioned rainwater discharge control system, it is better practice that the rainwater discharge control system additionally includes a predictive measure for predicting the amount of rainwater inflow into the pumping plant, and a calculation measure for calculating the number of operations of the rainwater pumping 'The rainwater pumps are required to discharge the rainwater inflows predicted by the forecasting measures. According to one of the inventions of claim 26, a rainwater discharge control method includes the steps of: determining a number of operations of the rainwater pump based on the predicted amount of rainwater inflow into one of the pumping plants; based on the pump One of the plant's water level changes 'controls the start and stop of the rainwater pump; and changes a start water level and a stop water level of the rainwater pump based on a water level in the sewage pipe near one of the predetermined upstream locations of the chestnut plant, To delay the start-up time of one of the rain pumps. According to one of the inventions described in claim 27, a rainwater-16-200923211 emission support method for predicting the inflow of water from a sewage pipe into a chestnut plant or a sewage plant, the rainwater drainage support method includes The following steps: measuring the rainfall of an area where the sewage pipe is located at a predetermined frequency; measuring the flow of water into the pump or sewage plant; and using a system identification method using a nonlinear Hammerstein model to present the current flow Based on some past flows, some past rainfall and some past rainfall indices to predict future water inflows. According to one of the inventions of claim 28, a rainwater emission control system is used for inflow of rainwater inflows predicted by using the rainwater drainage support method described in claim 7 The rainwater inflow of the pump plant determines the number of operations of the rainwater pump and controls the start and stop of the rainwater pump based on a change in the water level of the pumping plant. The rainwater discharge control system includes: a detection measure, Detecting a water level at a predetermined upstream location, wherein the water system flows to the pump factory; and a change measure for changing at least the rainwater pump based on the flow condition of the water level detected by the detection measure a predetermined start water level and one of a predetermined stop water level; and a calculation measure for calculating a number of operations of the rain water pump, the rain water pump being discharged from the rainwater discharge support method described in item 7 of the patent application scope The water that flows into the rainwater. According to the rainwater drainage support system and the rainwater drainage support method of the present invention, the linear mapping unit performs a linear mapping process, and the rainfall time series data matrix of the k X η matrix or the predicted rain time series data of the k χ η matrix The matrix is transformed into a linear mapping data matrix of a k X m matrix (m <n). Therefore, the number of variables input to the inflow prediction unit stream -17-200923211 into the volume prediction model can be reduced to promote a more accurate prediction of the rainwater inflow by the inflow prediction unit. According to another rainwater discharge support system and a rainwater discharge support method of the present invention, since the influent water quality is predicted by a system identification method based on the amount of influent water quality and past rainfall (non-linear Hammerstein model), it is easy to Execute the forecast. Since the accumulation and discharge of influent water can be properly controlled to be consistent with the predicted results, the adverse effects on the environmental side are reduced. According to the rainwater drainage support system and the rainwater drainage support method, one of the start water level and one stop water level of the rainwater pump is changed based on the predicted inflow amount of one of the inflowing pumps, and is consistent with the state of one water flow. Therefore, a discharge system can be stably operated. [Embodiment] FIG. 1 is a block diagram showing a rainwater discharge system according to an embodiment of the present invention. Fig. 2 is a block diagram showing a control system of a rainwater discharge system as shown in Fig. 1. As shown in Fig. 1, the rainwater discharge system of the present embodiment includes a rainfall amount measuring unit 10 for measuring the rainfall of a plurality of areas in a region where the sewage pipe 20 or other equivalent is located. A pumping plant including an inflow pipe 21 and a rainwater pumping well 30 is located downstream of the sewage pipe 20. The rainfall measurement unit 1 is composed of a radar rain gauge 11 and a plurality of ground rain gauges. In this case, the rainfall measurement unit 1 is composed of the radar rain gauge 11 , and a radar base station measures the total linearity of the radiation wave by using the n-block block -18-200923211. The same reflection characteristics occur when contacted with raindrops. As shown in Fig. 1, one area of rainfall measured by an observation network is divided into n-blocks, so that each block is composed of a square. Each side of the square has a length of several hundred meters to several kilometers. . Therefore, the rainfall divided into n pieces can be measured separately. When the rainfall measurement unit 10 is composed of the radar rain gauge 1 1 , when the rainfall measurement unit 10 is composed of a plurality of ground rain gauges 12 , the measured rainfall An area can be subdivided into a smaller observation network. Therefore, the number η of the regions measured by the ground rain gauge can be increased. For example, the sewage pipe 20 is a mixed type. The rainwater falling in a region where the sewage pipe 20 is located, together with the domestic wastewater in this area, flows together with the industrial wastewater through the sewage pipe 20. A pump factory is located downstream of the sewage pipe 20, and an inflow pipe 2 1 is located at the plant. The inflow pipe 2 1 temporarily collects rainwater coming in from the sewage pipe 20 . As shown in Fig. 1, a water quality meter 25, a flow meter 2, and a water level gauge 27 are located in the sewage pipe 20 and close to the inflow pipe 21. In the pump plant, the rainwater pump well 30 of the silt sedimentation tank is located downstream of the inflow pipe 21. The pump plant also has a sewage pump well 31 and a rainwater retention tank 32. The influent water passes through the inflow pipe 21 and the sedimentation sedimentation tank (not shown), from the sewage pipe 20 to the rainwater pump well 30 of the silt sedimentation tank, the sewage pump well 3 1 and the rainwater retention tank 32. The rainwater pump well 30 of the sedimentation tank typically has an overflow weir downstream of the inflow pipe 21. Regular influent water flows into the sewage pump well 3 1 . On the other hand, when a large amount of influent water including rainwater flows into the rainwater pump well 30 of the silt sedimentation tank during the rain, the water level of the inflow pipe 21 rises. -19- 200923211 Therefore, the water overflowing the high overflow weirs flows into the rainwater sediments of the silt sedimentation tank. Therefore, the rainwater retention tank 3 2 collects the water from the place and flows into the pump factory. The rainwater stagnation tank 3 2 receives the water flowing into the water through the overflow raft which is not shown in the figure. Here, it does not mean that one of the overflow rafts is located between the rainwater retention tank 3 2 and the flow tube 2 1 , or It is not indicated that one of the gates is located between the rainwater stagnation tank 32 and the inflow pipe 21. The rainwater pump well of the silt sedimentation tank includes a plurality of rainwater pumps 35. Into the sedimentation tank, the rainwater (rainwater) of the rainwater chestwater is released to a river 34 by the rainwater 35. The sewage pump well 31 includes a sewage pump 36. The influent (sewage) entering the sewage pump well 31 is collected here and sent to a sewage disposal facility through the sewage pump 36. The figure does not indicate that the sewage pump 36 is subject to a sewage treatment. The treated water is then released to a river and other equivalents. If the sewage exceeds the range that the sewage treatment equipment can handle, the sewage will be sent to the rainwater retention tank 3 2 to be concentrated to avoid the sewage overflowing the sewage pump well 31. The rainwater retention tank 3 2 includes a rainwater retention tank 3 7 . When the water flowing into the inflow pipe 21 is significantly reduced, the water accumulated in the rainwater retention tank 32 is sent back to the sewage through the rainwater retention tank pump 37. In other words, the sewage pipe 20 is discharged. The influent water passes through the inflow pipe 2 between the one end joint located in the sewage 2 淤 and the silt sedimentation tank. The figure does not indicate that the rainwater pump well 30, the sewage chestnut and the rainwater detention tank 32 are concentrated in the silt sedimentation tank. One or all of them are then placed in a predetermined discharge destination. The flow meter 2 3 and the water quality meter 24 are located in the inflow pipe 21. The well is retained in the residual flow system or the water pump sluice tube and 31 is released. -20- 200923211 The speedometer 2 3 series is measured and collected in the inlet water level of the inflow pipe 2 1 . The water 24 system measures the influent water quality flowing into the inflow pipe 2 1 . As shown in Fig. 1, the rainwater pump well of the silt sedimentation tank is 30 packs to measure the water level of the rainwater pump well of the silt sedimentation tank. The enthalpy measured in the flow rate meter 23 of the inflow pipe 21 and the flow rate meter 38 of the rainwater chestnut well of the sump. A control system for the stormwater drainage system is illustrated. As shown in Figure 2, the rainwater drainage system includes: an inflow amount of 40, using the current sequential rainfall in multiple areas, to the influent inflow of one of the pre-pump plants; one influent water quality] 50, borrowed Using the above-mentioned rainfall to predict the influent water quality flowing into the pumping plant (such as the inflow pipe 2 1 ); and the rainwater pump control loss using the inflow inflow prediction mechanism 40 to predict at least the influent inflow amount of the rainwater pump 3 5 One start water level and one stop water level - * The detailed structure of the 〇 inflow prediction mechanism 40, as shown in Figure 3 j. As shown in FIG. 3, the inflow prediction mechanism 40 is provided with a unit 41 and an inflow prediction unit 42, wherein the linear entropy performs a linear mapping process, and the linear mapping process is measured by the conversion unit 1 〇 The rainfall time-series moment inflow prediction unit 42 of the plurality of regions uses the mapping data by the linear mapping unit 41 to predict that the rainwater inflow matrix generating unit 43 flowing into a target region is connected to the linear mapping unit 4. 1, for the quality measurement including a flow rate at any time in the silt sediment prediction mechanism to measure the inflow, such as the page measurement mechanism, a target plant! System 60, to adjust the linear entropy 41 system rainfall of Figure 4 Measure the array and get it linear. The representation moment is made a table -21 - 200923211 No matrix 'provided to the linear mapping unit 41 as a linear mapping process. The model identification unit 44 is connected to the inflow amount prediction unit 42 for composing an inflow amount prediction model, and supplies it to the inflow amount prediction unit 42 as a prediction of the rainwater inflow amount. The linear mapping unit 41 obtains a rainfall time series data matrix X, wherein the rainfall time series data matrix X represents the time series rainfall of each region, and the time series rainfall of each region is measured by the rainfall measuring unit 1 〇 The rainfall in the complex region and the linear mapping unit 41 perform a linear mapping process to convert a plurality of variables of the rainfall time series data matrix X into data of less variables to obtain a linear mapping data matrix Y. Specifically, the linear mapping unit 4 1 performs an arithmetic operation of the following equation (1). γ = XP... Equation (1) (where X is a matrix of rainfall timing data k (discrete time) X n (number of blocks of measured rainfall), and P system - η X m matrix (m <n) representation matrix, which is generated by the representation matrix generation unit 43, as explained below) In the rainfall time series data matrix, the rainfall in each block at time t is equivalent to a column of vectors Xt (t = 1, ..., k). Among them, the current rainfall data is defined as xq +丨 in the case of Yiyicheng Xi ’q step (lSqgk-1). The linear entropy unit 4 1 performs the equation (1) calculation to obtain the linear entropy data matrix Y of the k X m matrix. The inflow amount prediction unit 42 uses an inflow amount prediction model composed of the following model identification unit 44 to predict the rainwater inflow amount, in which the linear entropy data obtained by the linear-22-200923211 mapping unit 41 is input to The inflow forecasting model. As shown in Fig. 3, when the inflow amount prediction unit 42 is predicting the amount of rainwater inflow, the time series inflow amount data should be taken into consideration. Among them, the inflow prediction model is a black box model, which is determined by the relationship between the measured rainfall data in the past and the inflow data corresponding to the rainfall data. Taking into account the linear regression data of the rainfall to date obtained by the linear mapping unit 41 and the sequential inflow data so far, the inflow prediction model has been based on the relationship between the rainfall data and the past inflow data. The inflow amount prediction unit 42 can predict the model based on the inflow amount to predict the future inflow of the rainwater. The representation matrix making unit 43 calculates a main element from the past rainfall timing data by a principal component analysis of one of the multivariate analysis to obtain a load matrix, and by using the characteristics of the resulting load matrix, the matrix is transformed into The matrix P is represented as a linear mapping process of the linear mapping unit 41. Specifically, the representation matrix P of the η X m matrix is produced by a combination of a plurality of intrinsic vector elements of a variation matrix of a past rainfall time series data matrix or a common variation matrix. The matrix making unit 43 is shown in full detail with reference to FIG. In step 111, the variation matrix or the common variation matrix is calculated from the past rainfall time series data, wherein the past rainfall time series data is the rainfall measured in a region of a block. When the past rainfall time series data is represented by the matrix X h of kh X η, the elements constituting the representation matrix P are a matrix of the matrix Xh or a -23-200923211 η-order eigenvector Pj of the common variation matrix Sh ( j = 1,...,n) line vector ρι,...,Pm(m <n). Note that the eigenvector Pj is used to satisfy; I and 2$...2 λ ... into the η condition inherently 値Λ j (;j = 1 , ..., η). In step I1, as shown in the following equation (2), the first main component to the mth main component are represented by a m-th order column vector ti of the matrix T of kh X m at discrete time t (i = 1, ..., k).
Xh = TP' + E……方程式（2) 其中E係由終止在第m 分量（m<n)之一主要分量所產生的一誤差，此時符號’表 示該矩陣的轉置。 然後如圖4所示，累積貢獻率a係從步驟1 2所計算 出來。累積貢獻率係表不一主要分量T擁有多少該原始矩 陣Xh之資訊的値。該累積貢獻率係由下面方程式（3)所計 算 …方程式⑶ iel 2=1 下一步’在步驟13中，選定固有向量使得該累積貢 獻率a係大於一預定臨限値。最後，在步驟ι4中，整理 所選定之固有向量以得到nxrn矩陣之表示矩陣P。 在此方法中’表示矩陣製作單元43係製作表示矩 陣’其中該表示矩陣係由一負載矩陣所形成，而該負載矩 陣係在過去降雨量時序資料矩陣Xh之一主要元素之分析 期間所得之。 模型辨識單元44係構成以過去大量之降雨量時序資 -24 - 200923211 料矩陣爲基礎之流入量預測模型，其中將降雨 矩陣之多個變數資料轉換成較少之變數資料， 量之降雨時序流入量資料。 用於本發明之流入量預測模型係一黑盒子 辨識單元44係使用系統辨識方法以模擬過去 資料之輸入/輸出關係以及雨水流入量時序資: 過去降雨量時序資料係受制於線性對映過程， 流入一目標廠，而該系統辨識方法例如係藉最 以找出係數參數。一模擬方法包括，譬如說， 經網路的方法，或者如下所述一 H a m m e r s t e i η : 進水水質預測機制5 0之詳細架構係參照g 之。如圖5所示，進水水質預測機制5 0包括 裝置5 2、一資料記憶裝置5 3、一預測裝置5 4 置55、一運作支援裝置56，以及一控制裝置 氣候模式預測裝置5 8可提供在進水水質預測機 資料獲取裝置52定期獲取由降雨量量測 測之降雨量（雷達雨量計資料 52a與地面两 5 2 b )、流入管水位5 2 c以得到進水水量、進水 與各個栗35到泵37的運作狀況有關之泵資訊 各個栗3 5到栗3 7的排出量5 2 f。資料記憶裝 期由資料獲取裝置52所得之資料以儲存各種責 栗3 5到栗3 7的根據非線性H a m m e r s t e i η 儲存於資料記憶裝置5 3的各種資料，以預測 進水水量。非線性Ham me rste in模型係以下面 量時序資料 以及過去大 模型。模型 降雨量時序 料’其中該 而該雨水係 小平方方法 一種使用神 方法。 團5以說明 一資料獲取 、一計算裝 5 7。此外， I 制 5 0。 【元1 0所量 I量計資料 水質 5 2 d、 52e ，以及 置53藉定 :料。 模型，使用 進水水質與 簡化之模擬 -25- 200923211 方程式來表示。換言之，從方程式（4)可得到進水水質’ 同時從方程式（5 )可得到進水水量。 所預測之進水水質=目前時間（t)的水質X α 1 +過去 時間（t-Ι)的水質χα 2 +過去時間（卜2)的水質xa 3 +… +過去時間（t-Ι)的降雨量X 0 1 +過去時間〇2)的 降雨量 X /3 2 +… + (過去時間（t-Ι)的降雨量）2 X r 1 + (過去時間（t-2) 的降雨量）2 X r 2 +… + (過去時間（t-i)的降雨量）3 X 6 1 + (過去時間（t-2) y的降雨量）3 x<5 2+… ............方程式（4) 所預測之進水水量=目前時間⑴的水量χα 11 +過去時 間（t-ι)的水量12 +過去時間（t-2)的水量13 +… +過去時間（t-ι)的降雨量X A 1 1 +過去時間（t-2)的 降雨量x/3 12 +… + (過去時間（t-υ的降雨量）2 xr η + (過去時間（t-2) 的降雨量）2 12 +— + (過去時間（t-υ的降雨量）3 χβ 11 + (過去時間（t_2) 的降雨量）3 x 5 12 +一 方程式（5) -26- 200923211 在方程式（4)與方程式（5)，t表示現在時刻、t-l表示 第一個過去時間，與t-2表示第二個過去時間。同時， α1，α2，α3，·_·，α11，α12,α13，.·_，β1，β2πβ11，β12,.··,γ1，γ2，·.·， γ11,γ12,...，δ1，δ2,..·，及δ11,δ12,…爲係數，而且這些値係 由各設備所決定，取決於污水管20與流入管21的尺寸或 特性。 換言之，進水水質與進水水量係以一預定非線性關係 跟隨著過去降雨量而變。因此，以現在水質、某些過去水 質、某些過去降雨量與某些過去降雨量之指數爲基礎，模 型方程式（4)可預測未來水質。同理，以現在水量、某些 過去水量、某些過去降雨量與某些過去降雨量之指數爲基 礎，模型方程式（5)可預測未來水量。 視預測裝置5 4所預測之進水水質與進水水量而定， 計算裝置5 5計算各個泵3 5到泵3 7的運作量（運作數目、 旋轉速度等等）。雨水泵3 5運作量的計算方法係如下詳盡 說明以解釋雨水泵控制機制60。運作支援裝置56藉由預 測裝置5 4所預測之水質與水量的使用，以產生與進水的 蓄與放、流入淤沙沉澱池之雨水泵井3 0、污水泵井3丨與 雨水滯留池3 2有關之運作/停止命令。控制裝置5 7根據 該運作命令將一控制命令輸出至一相對儀器。氣候模式預 測裝置5 8根據降雨量以預測如一般降雨模式或大雨模式 之氣候模式。計算裝置5 5所計算出的數値可與氣候模式 預測裝置5 8所預測之氣候模式相比對而予以變更。 雨水栗控制機制6 0的細部架構係參照圖6與以說明 -27- 200923211 之。 雨水泵控制機制6 0係一可規劃之裝置，譬如該裝置 可以是一可程式化邏輯控制器（PLC)、工作地點、個人電 腦’與微電腦。同樣地，雨水泵控制機制6 0可係由複數 之計算器所組成之計算機系統的其中一項功能。 如圖6所示，雨水泵控制機制60計算裝置包括運作 數目計算裝置62、啓動/停止水位變更裝置63，以及水位 計選定裝置64。 運作數目計算裝置62係根據流入量預測機制40所預 測之雨水流入量預測値Q，以計算出所需的泵運作數目， 使用至少在雨水流入量、泵井水位Η ρ與泵排出量Q ρ的預 測値Q其中之一做爲一輸入。 接著’啓動/停止水位變更裝置63即依據至少在污水 管之上游水位Η!，…，Ηη、預測泵之運作數目，以及泵井目 前水位Ηρ的其中之一 ’以變更泵之—預定啓動水位與一 停止水位。 水位計選定裝置64係選定與切換一水位計以作爲該 等泵之一啓動/停止水位變更之指標，以符合如雷達雨量 計1 1的降雨量量測單元1 0所量測一污物排放區之平面式 降雨分配。 雨水泵控制機制6 0發送如啓動/停止（開/關）的各種控 制信號到該等雨水泵3 5，以控制如馬達與閥的泵機器設 備。 接下來’本實施例的一運作將繼續上面討論而說明如 -28- 200923211 下。 k入量預測機制4 0的運作係參閱圖3與圖4予以說 明之。 首先’如圖1所示’將觀測網分成η片區塊的降雨量 分別係由降雨量量測單元1 〇所量測。 線性對映之表7Κ矩陣Ρ係由表示矩陣製作單元4 3 所製作。具體而言，如圖4所示’在步騾η*,表示矩 陣製作單兀43自過去降雨量時序資料以計算變異矩陣或 共同變異矩陣’以及計算該變異矩陣或共同變異矩陣的一 固有向量Pj(j = 1，…，η)與一相對之固有値又j(j = 1，…，η)’其中該過去降雨量時序資料係在—區塊的—^ 地區所量測之降雨量。 在步驟1 2中，累積貢獻率a係從該固有値λ」所計算 出。在步驟13中，係選定該等固有向量ρ』以使累積貢獻 率a大於一預定臨限値。最後’在步驟14中，係整理所 選定之固有向量Pj以得η X m矩陣（m<n)之表示矩陣p。 線性對映單元4 1得到一降雨量時序資料矩陣χ，其 中該降雨量時序資料矩陣X係表示各地區之時序降雨 量’而各地區之時序降雨量係來自降雨量量測單元1 〇所 量測複數地區之降雨量，以及線性對映單元4 1施行一線 性對映過程’將降雨量時序資料矩陣X之多個變數轉換 成較少變數之資料，以得到一線性對映資料矩陣γ。 因爲降雨量時序資料矩陣X係一' k X η之矩陣（k: 離散時間，η :所量測降雨之區塊數），以及由表示矩陣製 -29- 200923211 作單元43所產生之表示矩陣p係一 η X m矩陣（m<n)， 因此所得之線性對映資料矩陣Y係一 k x m之矩陣。 因此’對線性對映單元41而言，降雨量時序資料的 數目即可從k X η之數目降至k X m之數目。 模型辨識單元44係以過去降雨量時序資料爲基礎以 構成流入量預測模型，其中將降雨量時序資料矩陣之多個 變數資料轉換成較少之變數資料，與過去雨水之時序流入 量。 流入量預測單元4 2使用由模型辨識單元4 4所構成之 流入量預測模型’以預測雨水流入量，其中由該線性對映 單元4 1所得之線性對映資料係輸入至該流入量預測模 型。 接下來，進水水質預測機制5 0之運作係參照圖5以 說明之。 一般而言，降雨啓動時所出現之首次沖刷，係在污水 下水道所發生的問題。雖然首次沖刷發生在降雨啓動時， 但沉積在污水管之淤泥何時流出污水管以流到泵廠則仍然 未知。因爲進水水質預測機制5 0可預測包括降雨期間所 沉積淤泥的進水水質與水量’所以係可能以適當方式來處 理該首次沖刷。例如’當首次沖刷發生時’盡可能將包括 淚泥的雨水匯集在污水管或雨水停滯池裡。盡可能抑制該 首次沖刷以免流入河川’反而將其導入污水處理廠，以遵 循河川排放的相關法規。因此’可進行環境友善的相關處 置。 -30- 200923211 一種由進水水質預測機制5 0所預測污水水質的方 法，可從降雨量得到進水水質。進水水質預測機制5 0使 用降雨量或降雨強度作爲一輸入以輸出一進水水質（大腸 桿菌、化學需氧量、生化需氧量、磷、氮等等）’使得系 統辨識方法提供一進水水質預測値。換言之’進水水質預 測機制50係根據模型方程式（4) ’以進行一回歸計算與一 多層回歸計算，其中該回歸計算係使用現在進水水質所量 測之値與過去某些進水水質所量測之値，該多層回歸計算 係使用過去某些真實量測的降雨量之値。另外’加入過去 某些指數（例如二次方與三次方）以預測一進水水質，該進 水水質對降雨係呈現一非線性關係。 如上所述，既然可預測流入該流入管2 1之進水水 質，運作支援裝置5 6即可進行適當控制’該控制係關於 進水的蓄與放以符合一進水水質。另外，既然預測裝置 5 4透過模型方程式（5 )可預測流入該流入管2 1之進水水 量，因此使用該預測進水水量與該預測進水水質可執行更 適當的運作控制。 當污水廠之污水管20所在區域下雨時，雨水即流入 該污水管20。如上所述，雨水流經位於污水管20的一終 點接頭之流入管21以及圖中沒表示之淤沙沈澱池之後， 匯聚於淤沙沈澱池之雨水泵井3 0、污水泵井31與雨水滯 留池3 2。既然淤沙沈澱池之雨水泵井3 0比起污水泵井3 1 淤沙具有較高的溢流堰，故流入污水泵井3 1之經常性進 水藉由污水泵3 6將其排放至污水處置設備。接著在污水 -31 - 200923211 處置設備所處理過的水隨即排放至河川或其他相等物。 在降雨期間，當包括雨水的大量進水流入該流入管 2 1時，則流入管2 1的水位上升’因此該進水即流入 '游沙 沈澱池之雨水泵井3 0。透過雨水泵3 5將匯集浙沙沈源池 之雨水泵井3 0的雨水釋放至河川3 4 ° 然而，由於在降雨時因首次沖刷所產生的進水係包含 沉積於污水管20之污泥’將該進水’也就是將包含污泥 之惡質雨水，透過雨水泵3 5排放至河川3 4並非符合環境 的最佳作法。因此，藉由預測包括首次沖刷之進水水質與 基於降雨狀況之污水進水量，方可決定是否做該水之匯集 或排放以實施安全與環境友善的控制。 如上所述，如雷達雨量計11與複數之地面雨量計1 2 之降雨量量測單元1 0係在下雨時用以量測一降雨量。資 料獲取裝置52係獲取資料52a與52b，以及資料記憶裝 置5 3係紀錄相同資料。資料獲取裝置5 2亦獲取流入管水 位52c與進水水質52d，其中流入管水位52c係得自水位 計2 3所在之流入管2 1的水位，進水水質5 2 d係得自水質 量測計24所在之流入管21的水質。資料記憶裝置5 3係 記錄資料52c與52d。資料獲取裝置52係定期獲取這些 資料，以及資料記憶裝置5 3係分別紀錄這些定期獲取之 資料。 預測裝置54根據非線性Hammerstein模型方程式（4) 與（5) ’使用紀錄在資料記憶裝置53的降雨量與進水水質 之相關資料，以預測進水水質與進水水量。在非線性 -32- 200923211Xh = TP' + E... Equation (2) where E is an error produced by terminating the principal component of one of the mth components (m<n), where the symbol ' represents the transpose of the matrix. Then, as shown in Fig. 4, the cumulative contribution rate a is calculated from step 12. The cumulative contribution rate is a measure of how much the primary component T has information about the original matrix Xh. The cumulative contribution rate is calculated by the following equation (3) ... Equation (3) iel 2 = 1 Next 'In step 13, the inherent vector is selected such that the cumulative contribution a is greater than a predetermined threshold 値. Finally, in step ι4, the selected eigenvectors are collated to obtain a representation matrix P of the nxrn matrix. In this method, the 'representation matrix making unit 43 is made to represent a matrix' in which the matrix is formed by a load matrix which is obtained during the analysis of the main elements of one of the past rainfall time series data matrices Xh. The model identification unit 44 constitutes an inflow prediction model based on a large amount of rainfall timing in the past, in which a plurality of variable data of the rainfall matrix is converted into less variable data, and the rainfall timing inflow Quantity data. The inflow prediction model used in the present invention is a black box identification unit 44 that uses the system identification method to simulate the input/output relationship of past data and the rainwater inflow timing: the past rainfall timing data is subject to the linear mapping process. Flow into a target plant, and the system identification method, for example, borrows the most to find the coefficient parameters. A simulation method includes, for example, a network method, or a H a m m e r s t e i η : the detailed architecture of the influent water quality prediction mechanism 50 is referred to g. As shown in FIG. 5, the influent water quality prediction mechanism 50 includes a device 52, a data storage device 53, a prediction device 5, a operation support device 56, and a control device climate mode prediction device. The influent water quality prediction machine data acquisition device 52 is provided to regularly obtain the rainfall measured by the rainfall amount (the radar rain gauge data 52a and the ground two 5 2 b), and the inflow pipe water level 5 2 c to obtain the influent water amount and the influent water. The pump information relating to the operation of each pump 35 to the pump 37 is 5 2 f to the pump 3 5 to the pump 3 7 . The data memory is loaded with the data obtained by the data acquisition means 52 to store various data stored in the data memory device 5 according to the nonlinear H a m m e r s t e i η to predict the amount of water influent. The nonlinear Ham me rste in model is based on the following time series data and the past large model. The model rainfall timing material 'where the rainwater system is a small square method using a god method. Group 5 to explain a data acquisition, a calculation of equipment 5 7 . In addition, I system 5 0. [Yuan 1 0 quantity I meter data water 5 2 d, 52e, and set 53 debit: material. The model is expressed using the influent water quality and simplified simulation -25- 200923211 equation. In other words, the influent water quality can be obtained from equation (4) while the influent water amount can be obtained from equation (5). Predicted influent water quality = current time (t) water quality X α 1 + past time (t-Ι) water quality χα 2 + past time (b 2) water quality xa 3 +... + past time (t-Ι) The rainfall in X 0 1 +Past time 〇2) X /3 2 +... + (the rainfall in the past (t-Ι)) 2 X r 1 + (previous time (t-2)) ) 2 X r 2 +... + (rainfall in the past (ti)) 3 X 6 1 + (rainfall in the past (t-2) y) 3 x<5 2+... ....... ..... Equation (4) Predicted influent water = current time (1) of water χα 11 + past time (t-ι) of water 12 + past time (t-2) of water 13 +... + past time (t-ι)'s rainfall XA 1 1 +Lake time in the past (t-2)x/3 12 +... + (past time (t-υ) 2 xr η + (past time (t- 2) Rainfall) 2 12 +— + (past time (t-υ rainfall) 3 χβ 11 + (previous time (t_2) rainfall) 3 x 5 12 + one program (5) -26- 200923211 In equation (4) and equation (5), t represents the current time, tl represents the first past time, and t-2 represents the second past. Meanwhile, α1, α2, α3,·_·, α11, α12, α13, .·_, β1, β2πβ11, β12, .··, γ1, γ2, ···, γ11, γ12,..., Δ1, δ2, .., and δ11, δ12, ... are coefficients, and these enthalpy are determined by each device, depending on the size or characteristics of the sewage pipe 20 and the inflow pipe 21. In other words, the influent water quality and the influent water amount It is a predetermined nonlinear relationship that follows the past rainfall. Therefore, model equation (4) predicts the future based on current water quality, some past water quality, some past rainfall and some past rainfall indices. Water quality. Similarly, based on the current water quantity, some past water quantity, some past rainfall and some past rainfall index, the model equation (5) can predict the future water volume. Depending on the water quality and the amount of influent water, the calculation device 5 5 calculates the operation amount (number of operations, rotation speed, etc.) of each pump 35 to pump 37. The calculation method of the operation amount of the rain pump 35 is explained in detail below to explain the rain. Pump control mechanism 60. Operation support device 56 by predicting device 5 4 The predicted use of water and water to generate an operation/stop command related to the storage and release of influent water, the rainwater pump wells flowing into the sedimentation tanks, the sewage pump wells, and the rainwater retention tanks. The control device 57 outputs a control command to a relative instrument according to the operational command. The climate mode predictive device 58 is based on rainfall to predict a climate pattern such as a general rainfall pattern or a heavy rain pattern. The number calculated by the computing device 55 can be changed as compared to the climate mode predicted by the climate mode predicting device 58. The detailed structure of the rainwater control mechanism 60 is described with reference to Figure 6 and -27-200923211. The rainwater pump control mechanism is a programmable device, such as a programmable logic controller (PLC), a work site, a personal computer, and a microcomputer. Similarly, the rainwater pump control mechanism 60 can be one of the functions of a computer system comprised of a plurality of calculators. As shown in Fig. 6, the rainwater pump control mechanism 60 calculating means includes an operation number calculating means 62, a start/stop water level changing means 63, and a water level gauge selecting means 64. The operation number calculation means 62 predicts 値Q based on the rainwater inflow amount predicted by the inflow amount prediction mechanism 40 to calculate the required number of pump operations, using at least the rainwater inflow amount, the pump well water level ρ, and the pump discharge amount Q ρ. One of the predictions 値Q is used as an input. Then, the start/stop water level changing device 63 changes the pump-predetermined starting water level according to at least the water level upstream of the sewage pipe Η!,..., Ηη, the number of predicted pump operations, and one of the current water level ρ of the pumping well. With a stop water level. The water level gauge selection device 64 selects and switches a water level gauge as an indicator of the start/stop water level change of one of the pumps to meet the measurement of a pollutant discharge by the rainfall measuring unit 10 such as the radar rain gauge 1 1 . Plane rainfall distribution in the area. The rainwater pump control mechanism 60 sends various control signals such as start/stop (on/off) to the rainwater pumps 35 to control pumping equipment such as motors and valves. Next, an operation of this embodiment will continue as discussed above and explained as -28-200923211. The operation of the k-input prediction mechanism 40 is explained with reference to Figs. 3 and 4. First, the rainfall that divides the observation network into n-blocks as shown in Fig. 1 is measured by the rainfall measurement unit 1 分别. The linearly mapped table 7 Κ matrix is produced by the representation matrix making unit 4 3 . Specifically, as shown in FIG. 4, 'in step 骡*, indicating that the matrix is created from the past rainfall time series data to calculate the variation matrix or the common variation matrix' and calculating an inherent vector of the variation matrix or the common variation matrix Pj(j = 1,...,η) is a relative intrinsic 値 and j(j = 1,...,η)' where the past rainfall timing data is measured in the -^ region of the block. In step 12, the cumulative contribution rate a is calculated from the inherent 値λ". In step 13, the intrinsic vectors ρ are selected such that the cumulative contribution rate a is greater than a predetermined threshold 値. Finally, in step 14, the selected eigenvectors Pj are arranged to obtain a matrix p of η X m matrices (m<n). The linear mapping unit 4 1 obtains a rainfall time series data matrix χ, wherein the rainfall time series data matrix X represents the time series rainfall of each region' and the time series rainfall of each region is from the rainfall measuring unit 1 Measuring the rainfall in the complex region, and the linear mapping unit 4 1 performs a linear mapping process to convert the plurality of variables of the rainfall time series data matrix X into less variable data to obtain a linear mapping data matrix γ. Because the rainfall time series data matrix X is a matrix of 'k X η (k: discrete time, η: the number of blocks of measured rainfall), and the representation matrix generated by the representation matrix -29-200923211 as unit 43 p is a η X m matrix (m < n), so the resulting linear entropy data matrix Y is a matrix of k x m. Therefore, for the linear mapping unit 41, the number of rainfall time series data can be reduced from the number of k X η to the number of k X m . The model identification unit 44 is based on the past rainfall time series data to form an inflow prediction model in which a plurality of variable data of the rainfall time series data matrix is converted into less variable data and a sequential inflow of rainwater in the past. The inflow amount prediction unit 42 uses the inflow amount prediction model ' composed of the model identification unit 44 to predict the rainwater inflow amount, wherein the linear entropy data obtained by the linear mapping unit 41 is input to the inflow amount prediction model. . Next, the operation of the influent water quality prediction mechanism 50 is explained with reference to FIG. 5. In general, the first scouring that occurs when rainfall starts is a problem in sewage sewers. Although the first flush occurred at the start of the rain, it is still unknown when the sludge deposited in the sewer pipe flows out of the sewer pipe to the pump plant. Since the influent water quality prediction mechanism 50 predicts the influent water quality and quantity of the sludge deposited during the rain period, it is possible to handle the first flush in an appropriate manner. For example, 'when the first flush occurs,' collect rainwater that includes tears as much as possible into the sewer pipe or rainwater stagnation tank. Suppress the first flush as much as possible to prevent it from flowing into the river and instead introduce it into the sewage treatment plant to comply with the regulations on river discharge. Therefore, it is possible to carry out environmentally friendly related disposal. -30- 200923211 A method for predicting the quality of sewage from the influent water quality prediction mechanism, which can obtain the influent water quality from rainfall. Influent water quality prediction mechanism 50 uses rainfall or rainfall intensity as an input to output an influent water quality (E. coli, chemical oxygen demand, biochemical oxygen demand, phosphorus, nitrogen, etc.), so that the system identification method provides a Water quality prediction 値. In other words, the 'influent water quality prediction mechanism 50 is based on model equation (4)' to perform a regression calculation and a multi-layer regression calculation, wherein the regression calculation uses the current influent water quality and some past influent water quality. After measurement, the multi-layer regression calculation uses some of the true measured rainfall in the past. In addition, some indexes (such as quadratic and cubic) were added to predict the influent water quality, which has a nonlinear relationship with the rainfall system. As described above, since it is possible to predict the influent water flowing into the inflow pipe 2, the operation support device 56 can perform appropriate control. The control system is compatible with the influent water quality. Further, since the predicting means 54 predicts the amount of influent water flowing into the inflow pipe 2 through the model equation (5), it is possible to perform more appropriate operational control using the predicted influent amount and the predicted influent water quality. When the area of the sewage pipe 20 of the sewage plant is raining, the rainwater flows into the sewage pipe 20. As described above, after the rainwater flows through the inflow pipe 21 located at the end joint of the sewage pipe 20 and the silt sedimentation tank not shown in the figure, the rainwater pump well 30, the sewage pump well 31, and the rainwater accumulated in the silt sedimentation tank are accumulated. Pool 3 2. Since the rainwater pump well 30 of the silt sedimentation tank has a higher overflow weir than the sewage pump well 3 1 silt, the recirculating water flowing into the sewage pump well 31 is discharged to the sewage by the sewage pump 36. Dispose of equipment. The water treated in the sewage -31 - 200923211 disposal facility is then discharged to rivers or other equivalents. During the rain, when a large amount of influent water including rainwater flows into the inflow pipe 2, the water level of the inflow pipe 2 1 rises. Therefore, the inflow water flows into the rainwater pump well 30 of the sand sedimentation tank. The rainwater collected from the rainwater pump wells of the Shensha pond in Zhesha was released to the river 3 4 ° through the rainwater pump 35. However, the influent system generated by the first flushing during the rainfall includes the sludge deposited on the sewage pipe 20 It is not the best practice to comply with the environment that the influent water, that is, the toxic rainwater containing sludge, is discharged through the rainwater pump to the river. Therefore, by predicting the influent water quality including the first flush and the amount of wastewater inflow based on the rainfall condition, it is possible to decide whether to do the pooling or discharge of the water to implement safe and environmentally friendly controls. As described above, the rainfall measuring unit 10 such as the radar rain gauge 11 and the plurality of ground rain gauges 12 are used to measure a rainfall when it rains. The data acquisition device 52 acquires the materials 52a and 52b, and the data memory device 53 records the same data. The data acquisition device 52 also obtains the inflow pipe water level 52c and the influent water quality 52d, wherein the inflow pipe water level 52c is obtained from the water level of the inflow pipe 2 1 where the water level gauge 23 is located, and the influent water quality is 5 2 d from the water quality measurement. The water quality of the inflow pipe 21 where the meter 24 is located. The data storage device 5 3 records the data 52c and 52d. The data acquisition device 52 periodically acquires the data, and the data storage device 53 records the regularly acquired data. The predicting means 54 uses the data relating to the rainfall and the influent water quality recorded in the data memory device 53 based on the nonlinear Hammerstein model equations (4) and (5)' to predict the influent water quality and the influent water amount. In nonlinear -32- 200923211
Hammerstein模型中，應考量降雨量以及進水水質或進水 水量之間的非線性關係。當降雨量增加時，進水水質或進 水水量與降雨量係呈現一指數關係（例如二次方與三次 方），故上述設計可應用在進水水質與進水水量。 基於如上所述之進水水質的預測値，運作支援裝置 5 6用以決定雨水滯留池3 2之雨水匯集、包括流入管21 的污水匯集、一運作/停止與雨水泵3 5的運作數目，以及 一運作/停止與污水泵3 6的運作數目。例如，當預測之進 水水質比預定臨限値還要差時，則運作支援裝置5 6產生 該運作命令以要求設於該污水泵井3 1之污水泵3 6運作， 同時除非該降雨量係超過一較高之臨限値，否則暫停設於 該雨水泵井3 0之雨水泵3 5運作。當預測到進水水質已惡 化時，則加速污水處理設備（高階作用、標準觸發污泥作 用、氧化渠作用）的前向式運作。 雨水泵控制機制6 0之運作係參照圖6到圖1 1以說明 之。具體來說，雨水泵控制機制6 0之施行步驟係說明於 圖7流程圖。在步驟2 1中，流入量預測機制4 0係預測一 流入量。在步驟2 2中，根據步驟2 1的結果得以計算出雨 水泵35的運作數目。在已公開之日本專利編號2000-328642詳細說明運作數目計算裝置62。運作數目計算裝 置62的運作係簡短說明如下。 藉使用由流入量預測機制40所預測之流進泵廠之雨 水流入量作爲一輸入參數與淤沙沈澱池雨水泵井3 0所量 測之水位値，以預測並計算數分鐘後（也就是5分鐘）或數 -33- 200923211 十分鐘後（也就是30分鐘）合適的泵運作數目°運作數目 計算裝置62根據下面方程式（6)，使用由流入量預測機制 4〇在一預定時間（t+Ι)所輸出之流入量預測値Qr'(t+1)’ 以計算出泵運作數目。 N'(t+1) = Qr'(t+1) / R ............方程式（6) ’ 其中 Ν(〇 係一泵運作數目[數目]、R係一泵分級[m3/s] ’與Qr’（t+1) 係在一預定時間（t+1)之後的流入量預測値。 啓動/停止水位的變更係說明在步驟23 °如果需要加 上一泵到所需的泵運作數目以排放所預測流入量時’則啓 動/停止水位變更裝置63係變更栗的一·預疋啓動水位以加 速或延遲泵之啓動/停止。 如圖8所示，根據所計算的運作數目決定啓動一泵的 運作時，此泵的一預定啓動水位與一預定停止水位 Hoff則分別修正到Hon - △ Η與Haff - △ Η。 如已公開之日本專利編號2000-3 28642所述，△ Η的 真正値係該泵井目前水位Ηρ該泵預定啓動水位HQn之間 的水位差，也就是△ H = Hcn - Hp。 ί AH=f Ηρ,Η},-%Ηη, \ hit 方程式⑺ 另外’如方程式（7)所示，使用一泵井水位、污水管 上游地點一水位，與時間等變動作爲變數可決定函數f。 同樣的’如圖9所示’當決定複數之泵的一啓動水位 與一停止水位，則可移動該啓動水位與該停止水位。 -34- 200923211 不同於在已公開日本專利編號2000-328642所述之泵 的啓動/停止水位變更裝置，本發明的特色係使用一污水 管的一上游地點的一水位以變更泵的一啓動水位與一停止 水位。 回到圖7 ’步驟2 4係決定表示污水管上游地點的水 位上升指標，是否超出一預定臨限値。在步驟25，只有 當該指標係超出該臨限値時，降低該泵的啓動/停止水 位。 步驟 26係決定表示污水管上游地點的水位上升指 標’是否低於一預定臨限値。只有當步驟25所述該指標 係低於該臨限値時，則提升該泵的啓動/停止水位。當表 示污水管上游地點的水位上升指標係介於上限與下限之間 時’則該泵根據在步驟27先前所設定的一預定水位以啓 動/停止該泵。在步驟2 8，當泵井目前水位Hp達到所設定 水位，則從一泵控制器輸出雨水排放泵的開/關信號。 圖1〇係表示關於水位量測値與水位上升速度之水位 上升指目標一覽表。如圖1 0所示，水位量測値與水位上 升速度之間的關係係事先決定，且該關係可用來作爲決定 一水位與一水位上升速度是否超過該臨限値。 圖1 1係說明污水管之一上游地點與污水管之一下游 地點兩者在水位時序資料之比較圖。何時上游水位上升與 何時下游水位上升之間的時間差可視爲雨水流經污水管的 傳送期間。換言之，當觀測到污水管上游地點的水位上升 時，則下游地點的水位必定在一段相當於傳送期間的時間 -35- 200923211 後才上升。 如上所述，根據本實施例之雨水排放支援系統與雨水 排放支援方法，可完成下面的事情。也就是藉使用流入量 預測機制40，則線性對映單元4 1施行一線性對映過程， 將一 k X η矩陣之降雨量時序資料矩陣X轉換成一 k X m矩陣之線性對映資料矩陣Y (m<n)。因此，可減少輸入 至該流入量預測單元42之流入量預測模型之變數數目， 以促成該流入量預測單元對雨水流入量更精準之預測。 此外，在過去降雨量時序資料矩陣Xh，使用包含一 變異矩陣或一共同變異矩陣Sh的一固有向量元素之表示 矩陣P，則線性對映單元4 1可得到線性對映資料矩陣 Y。因此，線性對映資料矩陣Y之元素可轉換成線性獨立 時序資料，其中該等元素彼此間係無關連性。 另外，根據構成表示矩陣之固有向量的固有値所計算 出的累積貢獻率 a，係大於一預定之臨限値係較佳的作 法。因此，在極可能受制於該線性對映過程之前，可得到 該線性對映資料矩陣Y以防止該降雨量時序矩陣X的資 訊流失。In the Hammerstein model, the nonlinear relationship between rainfall and influent water quality or influent water should be considered. When the rainfall increases, the influent water quality or the influent water volume has an exponential relationship with the rainfall (for example, quadratic and cubic), so the above design can be applied to the influent water quality and the influent water volume. Based on the prediction of the influent water quality as described above, the operation support device 56 is used to determine the rainwater collection of the rainwater retention tank 32, the collection of sewage including the inflow pipe 21, the operation/stop and the number of operations of the rainwater pump 35, As well as a number of operations/stops with the operation of the sewage pump 3 6 . For example, when the predicted influent water quality is worse than the predetermined threshold, the operational support device 56 generates the operational command to request the sewage pump 36 installed in the sewage pump well 31 to operate, unless the rainfall The system exceeds a higher threshold, otherwise the rain pump 35 installed in the rainwater pump well is suspended. When it is predicted that the influent water quality has deteriorated, the forward operation of the sewage treatment equipment (high-order action, standard trigger sludge action, oxidation canal action) is accelerated. The operation of the rainwater pump control mechanism 60 is described with reference to Figs. 6 to 11. Specifically, the execution steps of the rainwater pump control mechanism 60 are illustrated in the flow chart of FIG. In step 21, the inflow prediction mechanism 40 predicts an inflow. In step 2 2, the number of operations of the rainwater pump 35 is calculated based on the result of the step 21. The operation number calculating means 62 is explained in detail in Japanese Patent Laid-Open No. 2000-328642. The operation of the operation number calculating means 62 is briefly described below. By using the inflow of rainwater from the pumping plant predicted by the inflow forecasting mechanism 40 as an input parameter and the water level measured by the rainwater pump well in the silt sedimentation tank, it is predicted and calculated after a few minutes (ie 5 minutes) or number -33- 200923211 After 10 minutes (that is, 30 minutes) the appropriate number of pump operations. The operation number calculating means 62 uses the inflow amount prediction mechanism 4 for a predetermined time according to the following equation (6). +Ι) The output inflow is predicted 値Qr'(t+1)' to calculate the number of pump operations. N'(t+1) = Qr'(t+1) / R ............ Equation (6) ' where Ν(〇系一泵操作数[number], R系一Pump classification [m3/s] 'and Qr'(t+1) are the inflow predictions after a predetermined time (t+1). The start/stop water level change is described in step 23 ° if needed When pumping to the required number of pump operations to discharge the predicted inflow amount, the start/stop water level changing device 63 changes the start water level of the pump to accelerate or delay the start/stop of the pump. As shown in FIG. When the operation of starting a pump is determined according to the calculated number of operations, a predetermined starting water level of the pump and a predetermined stop water level Hoff are respectively corrected to Hon - Δ Η and Haff - Δ Η. As disclosed in Japanese Patent No. 2000- According to 3 28642, the true △ Η is the water level difference between the current water level of the pump Ηρ and the predetermined starting water level HQn of the pump, that is, Δ H = Hcn - Hp. ί AH=f Ηρ,Η}, -%Ηη , \ hit Equation (7) In addition, as shown in equation (7), using a pump well water level, a water level upstream of the sewer pipe, and changes in time as variables can be determined The number f. The same 'as shown in Fig. 9', when determining a starting water level of a plurality of pumps and a stop water level, the starting water level and the stopping water level can be moved. -34- 200923211 Unlike the Japanese Patent No. 2000 which has been published The start/stop water level changing device of the pump described in paragraph 328642, the feature of the present invention is to use a water level at an upstream location of a sewage pipe to change a starting water level of the pump and a stop water level. Return to Figure 7 'Step 2 4 The system determines whether the water level rise indicator at the upstream location of the sewer pipe exceeds a predetermined threshold. In step 25, the start/stop water level of the pump is lowered only when the indicator exceeds the threshold. Step 26 determines Whether the water level rise indicator 'where the sewage pipe is upstream is lower than a predetermined threshold 値. Only when the indicator is lower than the threshold 步骤 in step 25, the pump starts/stops the water level. When the water level rise indicator of the location is between the upper and lower limits, then the pump starts/stops the pump according to a predetermined water level previously set in step 27. In step 2, when the pump When the water level Hp reaches the set water level, the on/off signal of the rainwater discharge pump is output from a pump controller. Fig. 1 shows the water level rise target target list for the water level measurement and the water level rising speed. It is shown that the relationship between the water level measurement and the rising speed of the water level is determined in advance, and the relationship can be used as a decision whether the water level and the rising rate of the water level exceed the threshold. Figure 1 1 shows the upstream location of the sewage pipe. A comparison chart of the water level timing data with the downstream location of one of the sewage pipes. The time difference between when the upstream water level rises and when the downstream water level rises can be regarded as the time during which the rainwater flows through the sewage pipe. In other words, when it is observed that the water level at the upstream point of the sewer pipe rises, the water level at the downstream point must rise after a period equivalent to -35-200923211 during the transfer period. As described above, according to the rainwater discharge support system and the rainwater discharge support method of the present embodiment, the following matters can be accomplished. That is, by using the inflow prediction mechanism 40, the linear mapping unit 4 1 performs a linear mapping process, and converts the rainfall time series data matrix X of a k X η matrix into a linear mapping data matrix Y of a k X m matrix. (m<n). Therefore, the number of variables of the inflow amount prediction model input to the inflow amount prediction unit 42 can be reduced to contribute to the more accurate prediction of the inflow amount prediction unit by the inflow amount prediction unit. Further, in the past rainfall timing data matrix Xh, using the representation matrix P of an intrinsic vector element including a variation matrix or a common variation matrix Sh, the linear mapping unit 4 1 can obtain a linear mapping data matrix Y. Therefore, the elements of the linear mapping data matrix Y can be converted into linear independent time series data, wherein the elements are independent of each other. Further, the cumulative contribution rate a calculated based on the inherent 値 constituting the eigenvector representing the matrix is preferably a method larger than a predetermined threshold. Therefore, the linear mapping data matrix Y can be obtained to prevent the loss of the information of the rainfall timing matrix X before it is highly likely to be subject to the linear mapping process.
此外，線性對映單元4 1藉執行一具有使用一表示矩 陣P之線性對映過程，以獲得該線性對映資料矩陣Y，其 中該表示矩陣P係由一負載矩陣所形成，而該負載矩陣係 在過去降雨量時序資料矩陣xh之一主要元素之分析期間 所得之。因此，在極可能受制於該線性對映過程之前，可 得到該線性對映資料矩陣Y以防止該降雨量時序矩陣X -36- 200923211 的資訊流失。在同一時間，線性對映資料矩陣γ之元素 可轉換成線性獨立時序資料，其中該等元素彼此間係無關 連性。 另外，模型辨識單元4 4係根據過去降雨量時序資料 矩陣，以構成流入量預測模型，其中將降雨量時序資料矩 陣之多個變數資料轉換成較少之變數資料，以及過去雨水 時序流入量。因此，可減少輸入至模型辨識單元44之過 去降雨量時序資料數目。因此’可促成流入量預測模型的 形成。另外，由於流入量預測模型係根據線性獨立降雨量 時序資料所形成’其中在該線性獨立降雨量時序資料的該 等元素彼此間係無關連性，故可得到高精準性之流入量預 測模型。 此外’根據本實施例之雨水排放支援系統與雨水排放 支援方法，進水水質預測機制5 0根據由雷達雨量計1 1與 地面雨量計1 2其中之一或兩者所量測雨量，可預測進水 水質與進水水量。因此，可預測首次沖刷的發生。換言 之’運作支援裝置5 6從所預測之進水流入量予以辨別出 首次沖刷，並產生運作命令以要求該首次沖刷匯聚在該雨 水滯留池3 2，以及當降雨量低於一預定値時，則要求設 於雨水滯留池3 2的泵3 7將所匯聚之首次沖刷水送回至污 水泵井3 1。例如，當首次沖刷發生時，與雨水滯留池3 2 相通之流入管2 1之圖上沒表示之閘門係打開著，所以該 首次沖刷水得以在首次階段即匯聚在雨水滯留池3 2。當 進水水量夠低時，也就是當雷達雨量計1 1與地面雨量計 -37- 200923211 1 2無法施行降雨量量測時，則透過雨水滯留池泵3 7將所 匯聚之首次沖刷水從雨水滯留池3 2送回至污水泵井31。 透過污水泵3 6將污水輸送至一污水處置設備，圖中並沒 表示，污水泵3 6係受制於一污水處置（高階作用、標準觸 發污泥作用、氧化渠作用）。接下來，被處置過的水即釋 放至河川3 4，因此降低對環境的不良效應。 另外，根據本實施例之雨水排放控制系統與雨水排放 控制方法，雨水泵控制機制60藉使用污水管的一水位點 作爲一指標，以變更該等泵的一啓動水位與一停止水位， 對流入污水管的雨水而言，到該水位必須花幾分鐘的時間 方能到達一泵廠。接下來例如，當雨水真正突然地流入淤 沙沈澱池之前，加快該等泵的啓動時間以抑制在淤沙沈澱 池雨水泵井的水位上升。反過來說，延遲雨水泵的啓動時 間以減少雨水泵的排放量，可降低一地區所接收到雨水之 水質惡化。 除了如所計算帶有較大誤差之流入量預測値作爲一指 標之外，在污水管之上游地點所真實量測之水位以及在淤 沙沈澱池雨水栗井所真實量測之水位亦可當作指標。因 此，可降低該等雨水泵錯誤運作的可能性。 本發明之雨水排放系統應不受限於上述的實施例，可 在此加入不同的改變。 參閱圖1 2以說明本發明之雨水排放系統的一替代實 施例。在圖12中，與在圖1到圖Π所示之相同部位係以 相同的參考標號來標示，故省略這些部位的細部說明。 -38 - 200923211 如圖1 2所示，在本實施例的雨水排放系統與在圖1 到圖1 1所示之雨水排放系統不同地方僅在於流入量預測 機制40p係設有一降雨量預測單元1 5，該降雨量預測單 元1 5位於降雨量量測單元1 〇與線性對映單元4 1之間。 其他的結構與圖1到圖1 1所示之本實施例係完全相同。 流入量預測機制40p係參閱圖1 2以說明之。 如圖1 2所示，降雨量預測單元1 5係連接至該降雨量 量測單元1 0，故由降雨量量測單元1 0所量測η個區塊之 時序降雨量資料係送至降雨量預測單元1 5。 降雨量預測單元1 5根據所量測降雨量時序資料，以 預測在各區塊時序的一未來降雨量。 然後，由降雨量預測單元1 5所預測之未來降雨量資 料係送至線性對映單元4 1 ’其中降雨量預測單元1 5係連 接至該線性對映單元4 1。線性對映單元4 1係施行一線性 對映過程，將所預測之降雨量時序資料矩陣Ζ之多個變數 轉換成較少變數之資料’以得到一線性對映資料矩陣γ。 如上所述，在流入量預測機制4 0 Ρ ’線性對映單元4 1 係施行一線性對映過程’將k χ η之所預測降雨量時序 資料矩陣Ζ之多個變數轉換成k x m之線性對映資料矩 陣Y(m<n)之較少變數資料。因此’可減少輸入至該流入 量預測單元4 2之流入量預測模型之變數數目’以促成流 入量預測單元4 2對雨水流入量之預測。 然後，本發明的另一實施例係參閱圖1 3以說明之。 如圖1 3所示，在本實施例的雨水排放系統與在圖1 -39- 200923211 到圖1 1所示實施例之雨水排放系統不同地方僅在於進水 水質預測機制50p之資料獲取裝置52p的組成與圖5所示 之資料獲取裝置5 2的組成有所不同。其他的結構與圖1 到圖1 1所示之本實施例係完全相同。 進水水質預測機制5 Op係參閱圖1 2如下說明之。 不同於圖5所示之實施例，在本實施例之雨水排放系 統，由位在污水管2 0之水質量測計2 5、流速計2 6，與水 位計27所量測之資料25a到資料27a，係由資料獲取裝 置5 2p在一預定頻率下所獲取。這些資料係記錄在資料記 憶裝置53。資料獲取裝置52p與資料記憶裝置53可得到 與記錄來自氣候資訊系統之氣候資訊，並沒在圖中表示。 除了流進流入管21的進水水質與進水量之外，預測 裝置5 4預測在污水管2 0的水質、在污水管2 0的水流 量’以及從污水管2 0流到流入管2 1的污泥量或沈積量。 換言之，預測裝置54藉使用模型方程式（4)，以現在 污水管20的水質、某些過去污水管20的水質、某些過去 降雨量與某些過去降雨量之指數爲基礎，以預測尙未流進 一流入管2 1在一所在點的水質，其中該所在點係指水質 量測計2 5在污水管2 0的位置。 預測裝置5 4藉使用模型方程式（5 )’以現在污水管2 0 的水流量 '某些過去污水管20的水流量、某些過去降雨 量與某些過去降雨量之指數爲基礎，同樣也能預測尙未流 進〜流入管21在一所在點的水流量，其中該所在點係指 流速計2 6在污水管的位置。 -40- 200923211 另外’預測裝置54係從資料記憶裝置5 3獲取晴天的 污水管2 0水位，以計算於該污水管之污泥或沈積物之位 準，以及根據雨天時在污水管2 0的水流量與水質，以預 測流進該流入管2 1的首次沖刷之未來水流量與水質。 在本實施例中，既然由位在污水管2 0之水質量測計 2 5、流速計2 6，與水位計2 7所量測之資料2 5 a到資料 27a，係在一預定頻率下所獲取與記錄，所以藉由獲取雷 達雨量計1 1與地面雨量計1 2其中之一或兩者所量測的雨 量，可預測在一所在點的水流量與水質，其中該所在點係 指該量測裝置在污水管2 0的位置。換言之，可預測尙未 流進一流入管2 1的水流量與水質。因此，係可能以更明 確與精確的方式來支援或控制雨水滯留池3 2的雨水收 集、雨水泵3 5的運作與停止，以及污水泵3 6的運作與停 止。 根據雷達雨量計資料52a、地面雨量計資料52b，或 如氣象資訊系統的氣象預測，以計算出連續數天的晴天 日。同時可量測在連續晴天曰之污水管20水位27a以及 污水管20之污泥或沈積物之位準。接著，根據在污水管 20在一雨天時的水質25a與水流量26，以預測流進一流 入管2 1的進水水量與水質。當接連發生好幾天晴時，沈 積物通常容易黏附並堆積在污水管20。當沈積物在雨天 時流入污水管20 ’則根據上面事先所預測之沈積量，可 精準預測首次沖刷的進水量與進水水質。 係可能計算出來以進水水質BOD-SS負載（在每單位 -41 - 200923211 處理廠之生物重量所提供進水有機污染物質之負載量）爲 基礎。 【圖式簡單說明】 本發明之實施例係參照所附之圖式予以如下說明。注 意本發明不受限於如下所述之實施例’反而包括不同的賦 予本發明技術上創意之實施例。 圖1係一描述在本發明之一貫施例之雨水排放系統之 架構圖； 圖2係一描述在如圖1所示之雨水排放系統之一控制 系統所組成之方塊圖； 圖3係一說明在如圖2所示之控制系統之流入量預測 機制所組成之方塊圖； 圖4係一說明在如圖3所示之流入量預測機制之一表 示矩陣製作單元運作之流程圖； 圖5係一說明在如圖2所示之控制系統之進水水質預 測機制所組成之方塊圖； 圖6係一說明在如圖2所示之控制系統之雨水泵控制 機制所組成之方塊圖； 圖7係一說明在如圖6所示之雨水泵控制機制運作之 流程圖； 圖8係一描繪在如圖6所示之雨水泵控制機制之雨水 泵之一啓動水位與一停止水位之變更圖； 圖9係一描繪在如圖6所示之雨水泵控制機制之複數 -42- 200923211 之雨水泵之一啓動水位與一停止水位之一偏移運作圖； 圖1 〇係一說明在如圖6所示之雨水泵控制機制之一 水位上升之指標描述一覽表； 圖1 1係在如圖6所示之雨水泵控制機制，比較污水 管之一上游地點與污水管之一下游地點兩者在水位時序資 料之比較圖； 圖1 2係一說明在本發明之雨水排放系統之流入量預 測機制所另外組成之方塊圖；以及 圖1 3係一說明在本發明之雨水排放系統之雨水栗控 制機制所另外組成之方塊圖。 【主要元件符號說明】 1 0 :降雨量量測單元 1 1 :雷達雨量計 1 2 :地面雨量計 1 5 :降雨量預測單元 2〇 :污水管 2 1 :流入管 23 :流速計 2 3 :水位計 24 :水質量測計 25 :水質量測計 2 6 ·流速計 27 :水位計 43- 200923211 3 0 ·雨水栗井 3 1 :污水泵井 3 2 :雨水滯留池 34 :河川 3 5 :雨水泵 3 6 :污水栗 3 7 :雨水滯留池泵 3 8 :流速g十 40 :流入量預測機制 4〇p :流入量預測機制 41 :線性對映單元 42 :流入量預測機制 43 :表示矩陣製作單元 44 :模型辨識單元 5 0 :進水水質預測機制 -44 -In addition, the linear mapping unit 4 1 performs a linear mapping process using a representation matrix P to obtain the linear mapping data matrix Y, wherein the representation matrix P is formed by a load matrix, and the load matrix It is obtained during the analysis of the main elements of one of the past rainfall time series data matrix xh. Therefore, the linear mapping data matrix Y can be obtained to prevent the loss of information of the rainfall timing matrix X-36-200923211 before it is highly likely to be subject to the linear mapping process. At the same time, the elements of the linear mapping data matrix γ can be converted into linear independent time series data, where the elements are independent of each other. In addition, the model identification unit 44 is based on the past rainfall time series data matrix to form an inflow prediction model in which a plurality of variable data of the rainfall time series data matrix are converted into less variable data, and past rainwater timing inflows. Therefore, the number of past rainfall timing data input to the model identification unit 44 can be reduced. Therefore, it can contribute to the formation of the inflow prediction model. In addition, since the inflow prediction model is formed based on the linear independent rainfall time series data, in which the elements of the linear independent rainfall time series data are independent of each other, a highly accurate inflow prediction model can be obtained. Further, according to the rainwater drainage support system and the rainwater drainage support method of the present embodiment, the influent water quality prediction mechanism 50 is predictable based on the amount of rainfall measured by one or both of the radar rain gauge 1 1 and the ground rain gauge 1 2 Influent water quality and influent water volume. Therefore, the occurrence of the first flush can be predicted. In other words, the 'operation support device 566 identifies the first flush from the predicted influent inflow amount, and generates an operational command to request the first flush to converge in the rainwater retention tank 32, and when the rainfall is below a predetermined threshold, The pump 37 provided in the rainwater retention tank 32 is required to return the first flushed water to the sewage pump well 31. For example, when the first flush occurs, the gate that is not shown on the map of the inflow pipe 2 that communicates with the rainwater retention tank 3 2 is opened, so that the first flushing water is concentrated in the rainwater retention tank 32 in the first stage. When the amount of influent water is low enough, that is, when the radar rain gauge 1 1 and the ground rain gauge -37- 200923211 1 2 cannot perform the rainfall measurement, the first flushing water collected by the rainwater retention tank pump 37 is The rainwater retention tank 3 2 is returned to the sewage pump well 31. The sewage pump is transported to a sewage disposal facility through a sewage pump 36. The figure does not indicate that the sewage pump 36 is subject to a sewage treatment (high-order action, standard trigger sludge action, oxidation canal action). Next, the treated water is released to the river 34, thus reducing the adverse effects on the environment. In addition, according to the rainwater discharge control system and the rainwater discharge control method of the present embodiment, the rainwater pump control mechanism 60 uses a water level of the sewage pipe as an indicator to change a starting water level of the pumps and a stop water level, In the case of rainwater from sewage pipes, it takes a few minutes to reach a pumping plant. Next, for example, before the rainwater suddenly flows into the sedimentation tank, the start-up time of the pumps is accelerated to suppress the rise in the water level of the rainwater pump well in the sedimentation tank. Conversely, delaying the start-up time of the rainwater pump to reduce the discharge of the rainwater pump can reduce the deterioration of the water quality of the rainwater received in a region. In addition to the calculated inflow forecast with large errors as an indicator, the water level actually measured at the upstream location of the sewer pipe and the water level actually measured by the rainwater chestnut in the silt sedimentation tank can also be used. As an indicator. Therefore, the possibility of malfunction of such rainwater pumps can be reduced. The rainwater drainage system of the present invention should not be limited to the above-described embodiments, and various changes may be added thereto. Referring to Figure 12, an alternative embodiment of the stormwater drainage system of the present invention is illustrated. In Fig. 12, the same portions as those shown in Fig. 1 through Fig. 1 are denoted by the same reference numerals, and the detailed description of these portions will be omitted. -38 - 200923211 As shown in Fig. 12, the rainwater discharge system of the present embodiment is different from the rainwater discharge system shown in Figs. 1 to 11 only in that the inflow amount prediction mechanism 40p is provided with a rainfall prediction unit 1 5. The rainfall prediction unit 15 is located between the rainfall measuring unit 1 〇 and the linear mapping unit 41. The other structure is exactly the same as the embodiment shown in Figs. 1 to 11. The inflow prediction mechanism 40p is described with reference to FIG. As shown in FIG. 12, the rainfall prediction unit 15 is connected to the rainfall measuring unit 10, so the time-series rainfall data of the n blocks measured by the rainfall measuring unit 10 is sent to the rainfall. The quantity prediction unit 15 is. The rainfall prediction unit 15 5 predicts a future rainfall at each block timing based on the measured rainfall timing data. Then, the future rainfall information predicted by the rainfall prediction unit 15 is sent to the linear mapping unit 4 1 ' where the rainfall prediction unit 15 is connected to the linear mapping unit 4 1 . The linear entropy unit 4 1 performs a linear entropy process, converting a plurality of variables of the predicted rainfall time series data matrix into data of less variables to obtain a linear entropy data matrix γ. As described above, in the inflow prediction mechanism 4 0 Ρ 'linear mapping unit 4 1 performs a linear mapping process 'converts a plurality of variables of the predicted rainfall time series data matrix of k χ η into a linear pair of kxm The less variable data of the data matrix Y (m < n). Therefore, the number of variables of the inflow amount prediction model input to the inflow amount prediction unit 42 can be reduced to contribute to the prediction of the rainwater inflow amount by the inflow amount prediction unit 42. Then, another embodiment of the present invention is described with reference to FIG. As shown in FIG. 13 , the rainwater drainage system of the present embodiment differs from the rainwater drainage system of the embodiment shown in FIG. 1-39-200923211 to FIG. 11 only in the data acquisition device 52p of the influent water quality prediction mechanism 50p. The composition is different from the composition of the data acquisition device 52 shown in FIG. The other structure is exactly the same as that of the embodiment shown in Figs. 1 to 11. The influent water quality prediction mechanism 5 Op is described below with reference to Figure 1 2. Different from the embodiment shown in FIG. 5, in the rainwater discharge system of the present embodiment, the data 25a measured by the water quality meter 25 of the sewage pipe 20, the flow rate meter 2, and the water level gauge 27 are The data 27a is acquired by the data acquisition device 52p at a predetermined frequency. These data are recorded in the data memory device 53. The data acquisition means 52p and the data storage means 53 can obtain and record climate information from the climate information system, and are not shown in the figure. In addition to the influent water quality and the influent amount flowing into the inflow pipe 21, the predicting device 54 predicts the water quality in the sewage pipe 20, the water flow rate in the sewage pipe 20, and the flow from the sewage pipe 20 to the inflow pipe 2 1 The amount or amount of sludge deposited. In other words, the predicting device 54 uses the model equation (4) to predict the water quality of the present sewage pipe 20, the water quality of some of the past sewage pipes 20, some past rainfall and some past rainfall indices. The water flowing into an inflow pipe 2 1 at a point where the point refers to the position of the water quality meter 2 at the sewage pipe 20. The predictive device 54 is based on the use of model equation (5) 'based on the current flow of the sewage pipe 20', some of the past water pipe 20 water flow, some past rainfall and some past rainfall indices, It is possible to predict the flow of water that does not flow into the inflow pipe 21 at a point where the point is the position of the flow meter 26 at the sewage pipe. -40- 200923211 In addition, the 'predictive device 54 obtains the 20-water level of the sewage pipe from the data memory device 53 to calculate the level of the sludge or sediment in the sewage pipe, and according to the rainy day in the sewage pipe 2 0 The water flow and water quality are predicted to flow into the inflow pipe 21 for the first flush of the future water flow and water quality. In the present embodiment, since the data measured by the water quality meter 2 5 of the sewage pipe 20, the flow rate meter 2 6, and the water level gauge 27 are 2 5 a to the data 27a, at a predetermined frequency. Acquisition and recording, so by obtaining the rainfall measured by one or both of the radar rain gauge 1 1 and the ground rain gauge 12 , the water flow and water quality at a point can be predicted, where the point is The measuring device is at the position of the sewage pipe 20. In other words, it is possible to predict the flow of water and water quality that does not flow into an inflow pipe 2 1 . Therefore, it is possible to support or control the rainwater collection of the rainwater retention tank 32, the operation and stop of the rainwater pump 35, and the operation and stop of the sewage pump 36 in a more precise and precise manner. Based on radar rain gauge data 52a, ground rain gauge data 52b, or meteorological forecasts such as weather information systems, a few days of sunny days are calculated. At the same time, the level of the water level 27a of the sewage pipe 20 and the sludge or sediment of the sewage pipe 20 in a continuous sunny day can be measured. Next, based on the water quality 25a and the water flow rate 26 in the sewage pipe 20 on a rainy day, the amount of influent water and water quality flowing into the first-class inlet pipe 21 are predicted. When it occurs several days in a row, the deposits are usually easily adhered and accumulated in the sewage pipe 20. When the sediment flows into the sewer pipe 20' on a rainy day, the amount of water influent and the influent water quality of the first flush can be accurately predicted based on the amount of sediment previously predicted. It may be calculated based on the BOD-SS loading of the influent water (the amount of influent organic pollutants supplied by the biological weight of the treatment plant per unit -41 - 200923211). BRIEF DESCRIPTION OF THE DRAWINGS Embodiments of the present invention are described below with reference to the accompanying drawings. It is to be noted that the invention is not limited to the embodiments described below, but instead includes different embodiments that give the inventive idea of the invention. BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 is a block diagram showing a rainwater discharge system in accordance with a consistent embodiment of the present invention; Figure 2 is a block diagram showing a control system of a rainwater discharge system as shown in Figure 1; A block diagram of the inflow prediction mechanism of the control system as shown in FIG. 2; FIG. 4 is a flow chart showing the operation of the matrix fabrication unit in one of the inflow prediction mechanisms shown in FIG. 3; 1 is a block diagram showing the influent water quality prediction mechanism of the control system shown in FIG. 2; FIG. 6 is a block diagram showing the control mechanism of the rainwater pump in the control system shown in FIG. 2; 1 is a flow chart illustrating the operation of the rainwater pump control mechanism as shown in FIG. 6. FIG. 8 is a diagram showing a change of a starting water level and a stop water level of a rainwater pump of the rainwater pump control mechanism shown in FIG. Figure 9 is a diagram showing the operation of one of the start water level and one stop water level of the rain pump of the plurality of rain pump control mechanisms shown in Figure 6 as shown in Figure 6. Figure 1 shows a diagram of Figure 1 Rain pump control mechanism as shown A description of the indicators for the rise in water level; Figure 1 is a comparison of the timing data of the rainwater pump control mechanism shown in Figure 6, comparing the upstream location of one of the sewer pipes with the downstream location of one of the sewer pipes; Figure 1 2 BRIEF DESCRIPTION OF THE DRAWINGS FIG. 13 is a block diagram showing the additional composition of the inflow prediction mechanism of the rainwater discharge system of the present invention; and FIG. 13 is a block diagram showing the additional composition of the rainwater control mechanism of the rainwater discharge system of the present invention. [Main component symbol description] 1 0 : Rainfall measurement unit 1 1 : Radar rain gauge 1 2 : Ground rain gauge 1 5 : Rainfall prediction unit 2: Sewer pipe 2 1 : Inflow pipe 23 : Flow rate meter 2 3 : Water level gauge 24: Water quality meter 25: Water quality meter 2 6 · Flow meter 27: Water level gauge 43- 200923211 3 0 · Rainwater chestnut 3 1 : Sewage pump well 3 2: Rainwater retention tank 34: River 3 5 : Rainwater pump 3 6 : Sewage pump 3 7 : Rainwater retention tank pump 3 8 : Flow rate g 10 40 : Inflow prediction mechanism 4〇p : Inflow prediction mechanism 41 : Linear mapping unit 42 : Inflow prediction mechanism 43 : Representation matrix Production unit 44: model identification unit 5 0: influent water quality prediction mechanism - 44 -